[dpdk-dev] [PATCH 0/3] disable vector PMD for i686

[dpdk-dev] [PATCH 0/3] disable vector PMD for i686

[Qi Zhang]

Vector PMD does not support i686, it should not be compiled with i686 arch.
Though there is no compile error and sometimes it does works, but it is not
designed for i686 orignally.
We saw some test failed when vPMD is invovled on i686, so it is necessary to
fix.

Qi Zhang (3):
  net/i40e: disable vector PMD for i686
  net/ixgbe: disable vector PMD for i686
  net/fm10k: disable vector PMD for i686

 drivers/net/fm10k/Makefile | 2 ++
 drivers/net/i40e/Makefile  | 2 +-
 drivers/net/ixgbe/Makefile | 2 +-
 3 files changed, 4 insertions(+), 2 deletions(-)

-- 
2.9.3

[dpdk-dev] [PATCH 1/3] net/i40e: disable vector PMD for i686

[Qi Zhang]

Vector PMD does not support i686, it should not be compiled

Fixes: 8e109464c02292 ("i40e: allow vector Rx and Tx usage")
Cc: stable@dpdk.org

Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
---
 drivers/net/i40e/Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/i40e/Makefile b/drivers/net/i40e/Makefile
index 56f210d..0a3824e 100644
--- a/drivers/net/i40e/Makefile
+++ b/drivers/net/i40e/Makefile
@@ -101,7 +101,7 @@ ifeq ($(CONFIG_RTE_ARCH_ARM64),y)
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_INC_VECTOR) += i40e_rxtx_vec_neon.c
 else ifeq ($(CONFIG_RTE_ARCH_PPC_64),y)
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_INC_VECTOR) += i40e_rxtx_vec_altivec.c
-else
+else ifeq ($(CONFIG_RTE_ARCH_X86_64),y)
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_INC_VECTOR) += i40e_rxtx_vec_sse.c
 endif
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_PMD) += i40e_ethdev_vf.c
-- 
2.9.3

[dpdk-dev] [PATCH 2/3] net/ixgbe: disable vector PMD for i686

[Qi Zhang]

Vector PMD does not support i686, so do not compile it.

Fixes: 429c6d86b371 ("ixgbe: prepare for vector pmd")
Cc: stable@dpdk.org

Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
---
 drivers/net/ixgbe/Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ixgbe/Makefile b/drivers/net/ixgbe/Makefile
index 0a6b7f2..8261639 100644
--- a/drivers/net/ixgbe/Makefile
+++ b/drivers/net/ixgbe/Makefile
@@ -112,7 +112,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_IXGBE_PMD) += ixgbe_pf.c
 SRCS-$(CONFIG_RTE_LIBRTE_IXGBE_PMD) += ixgbe_flow.c
 ifeq ($(CONFIG_RTE_ARCH_ARM64),y)
 SRCS-$(CONFIG_RTE_IXGBE_INC_VECTOR) += ixgbe_rxtx_vec_neon.c
-else
+else ifeq ($(CONFIG_RTE_ARCH_X86_64),y)
 SRCS-$(CONFIG_RTE_IXGBE_INC_VECTOR) += ixgbe_rxtx_vec_sse.c
 endif
 
-- 
2.9.3

[dpdk-dev] [PATCH 3/3] net/fm10k: disable vector PMD for i686

[Qi Zhang]

Vector PMD does not support i686, it should not be compiled.

Fixes: a6ce64a97520 ("fm10k: introduce vector driver")
Cc: stable@dpdk.org

Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
---
 drivers/net/fm10k/Makefile | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/fm10k/Makefile b/drivers/net/fm10k/Makefile
index a8e8136..458ad66 100644
--- a/drivers/net/fm10k/Makefile
+++ b/drivers/net/fm10k/Makefile
@@ -95,6 +95,8 @@ SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_common.c
 SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_mbx.c
 SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_vf.c
 SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_api.c
+ifeq ($(CONFIG_RTE_ARCH_X86_64),y)
 SRCS-$(CONFIG_RTE_LIBRTE_FM10K_INC_VECTOR) += fm10k_rxtx_vec.c
+endif
 
 include $(RTE_SDK)/mk/rte.lib.mk
-- 
2.9.3

[dpdk-dev] [PATCH 0/3 v2] disable vector PMD for i686

[Qi Zhang]

Vector PMD is not designed for i686 orginally, but it still can be active
with i686 compile option.
Below are observed failure when vPMD is invovled on i686
(but may not limited to)

1) memory overwrite when assign 2 mbuf points to rx return points.
_mm_storeu_si128((__m128i *)&rx_pkts[pos+2], mbp2)

2) rearm_data is not 16 bytes aligned that cause general-protection exception
_mm_store_si128((__m128i *)&rx_pkts[0]->rearm_data, rearm0);

So the patch set will exclude Vector PMD from compile with i686 configure.

Qi Zhang (3):
  net/i40e: disable vector PMD for i686
  net/ixgbe: disable vector PMD for i686
  net/fm10k: disable vector PMD for i686

 drivers/net/fm10k/Makefile | 2 ++
 drivers/net/i40e/Makefile  | 2 +-
 drivers/net/ixgbe/Makefile | 2 +-
 3 files changed, 4 insertions(+), 2 deletions(-)

-- 
2.9.3

[dpdk-dev] [PATCH 1/3 v2] net/i40e: disable vector PMD for i686

[Qi Zhang]

Vector PMD does not support i686, it should not be compiled

Fixes: 8e109464c02292 ("i40e: allow vector Rx and Tx usage")
Cc: stable@dpdk.org

Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
---
 drivers/net/i40e/Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/i40e/Makefile b/drivers/net/i40e/Makefile
index 56f210d..0a3824e 100644
--- a/drivers/net/i40e/Makefile
+++ b/drivers/net/i40e/Makefile
@@ -101,7 +101,7 @@ ifeq ($(CONFIG_RTE_ARCH_ARM64),y)
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_INC_VECTOR) += i40e_rxtx_vec_neon.c
 else ifeq ($(CONFIG_RTE_ARCH_PPC_64),y)
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_INC_VECTOR) += i40e_rxtx_vec_altivec.c
-else
+else ifeq ($(CONFIG_RTE_ARCH_X86_64),y)
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_INC_VECTOR) += i40e_rxtx_vec_sse.c
 endif
 SRCS-$(CONFIG_RTE_LIBRTE_I40E_PMD) += i40e_ethdev_vf.c
-- 
2.9.3

[dpdk-dev] [PATCH 2/3 v2] net/ixgbe: disable vector PMD for i686

[Qi Zhang]

Vector PMD does not support i686, so do not compile it.

Fixes: 429c6d86b371 ("ixgbe: prepare for vector pmd")
Cc: stable@dpdk.org

Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
---
 drivers/net/ixgbe/Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ixgbe/Makefile b/drivers/net/ixgbe/Makefile
index 0a6b7f2..8261639 100644
--- a/drivers/net/ixgbe/Makefile
+++ b/drivers/net/ixgbe/Makefile
@@ -112,7 +112,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_IXGBE_PMD) += ixgbe_pf.c
 SRCS-$(CONFIG_RTE_LIBRTE_IXGBE_PMD) += ixgbe_flow.c
 ifeq ($(CONFIG_RTE_ARCH_ARM64),y)
 SRCS-$(CONFIG_RTE_IXGBE_INC_VECTOR) += ixgbe_rxtx_vec_neon.c
-else
+else ifeq ($(CONFIG_RTE_ARCH_X86_64),y)
 SRCS-$(CONFIG_RTE_IXGBE_INC_VECTOR) += ixgbe_rxtx_vec_sse.c
 endif
 
-- 
2.9.3

[dpdk-dev] [PATCH 3/3 v2] net/fm10k: disable vector PMD for i686

[Qi Zhang]

Vector PMD does not support i686, it should not be compiled.

Fixes: a6ce64a97520 ("fm10k: introduce vector driver")
Cc: stable@dpdk.org

Signed-off-by: Qi Zhang <qi.z.zhang@intel.com>
---
 drivers/net/fm10k/Makefile | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/fm10k/Makefile b/drivers/net/fm10k/Makefile
index a8e8136..458ad66 100644
--- a/drivers/net/fm10k/Makefile
+++ b/drivers/net/fm10k/Makefile
@@ -95,6 +95,8 @@ SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_common.c
 SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_mbx.c
 SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_vf.c
 SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_api.c
+ifeq ($(CONFIG_RTE_ARCH_X86_64),y)
 SRCS-$(CONFIG_RTE_LIBRTE_FM10K_INC_VECTOR) += fm10k_rxtx_vec.c
+endif
 
 include $(RTE_SDK)/mk/rte.lib.mk
-- 
2.9.3

Re: [dpdk-dev] [PATCH 0/3 v2] disable vector PMD for i686

[Bruce Richardson]

On Thu, Apr 27, 2017 at 03:01:04AM -0400, Qi Zhang wrote:
> Vector PMD is not designed for i686 orginally, but it still can be active
> with i686 compile option.
> Below are observed failure when vPMD is invovled on i686
> (but may not limited to)
> 
> 1) memory overwrite when assign 2 mbuf points to rx return points.
> _mm_storeu_si128((__m128i *)&rx_pkts[pos+2], mbp2)

Is this a serious issue that prevents us using the driver? I think it's been
in the code for quite some time. Can it not be relatively easily fixed for
32-bit builds?

> 
> 2) rearm_data is not 16 bytes aligned that cause general-protection exception
> _mm_store_si128((__m128i *)&rx_pkts[0]->rearm_data, rearm0);
> 

Good catch. I think this is also an easy fix. My preferred fix is to
explicitly align the rearm data on a 16-byte boundary. It would add some
padding to the middle of cacheline0 of the mbuf, but given that we
explicitly move other data to cacheline1, we will have padding on 32-bit
anyway, be it in the middle or the end of the mbuf cachelines.

> So the patch set will exclude Vector PMD from compile with i686 configure.
> 
> Qi Zhang (3):
>   net/i40e: disable vector PMD for i686
>   net/ixgbe: disable vector PMD for i686
>   net/fm10k: disable vector PMD for i686
> 
>  drivers/net/fm10k/Makefile | 2 ++
>  drivers/net/i40e/Makefile  | 2 +-
>  drivers/net/ixgbe/Makefile | 2 +-
>  3 files changed, 4 insertions(+), 2 deletions(-)
> 
> -- 
> 2.9.3
> 

Re: [dpdk-dev] [PATCH 0/3 v2] disable vector PMD for i686

[Thomas Monjalon]

27/04/2017 11:25, Bruce Richardson:
> On Thu, Apr 27, 2017 at 03:01:04AM -0400, Qi Zhang wrote:
> > Vector PMD is not designed for i686 orginally, but it still can be active
> > with i686 compile option.
> > Below are observed failure when vPMD is invovled on i686
> > (but may not limited to)
> > 
> > 1) memory overwrite when assign 2 mbuf points to rx return points.
> > _mm_storeu_si128((__m128i *)&rx_pkts[pos+2], mbp2)
> 
> Is this a serious issue that prevents us using the driver? I think it's been
> in the code for quite some time. Can it not be relatively easily fixed for
> 32-bit builds?
> 
> > 
> > 2) rearm_data is not 16 bytes aligned that cause general-protection exception
> > _mm_store_si128((__m128i *)&rx_pkts[0]->rearm_data, rearm0);
> > 
> 
> Good catch. I think this is also an easy fix. My preferred fix is to
> explicitly align the rearm data on a 16-byte boundary. It would add some
> padding to the middle of cacheline0 of the mbuf, but given that we
> explicitly move other data to cacheline1, we will have padding on 32-bit
> anyway, be it in the middle or the end of the mbuf cachelines.
> 
> > So the patch set will exclude Vector PMD from compile with i686 configure.

Please try to fix the drivers instead of turning them off.

Re: [dpdk-dev] [PATCH 0/3 v2] disable vector PMD for i686

[Bruce Richardson]

On Fri, Apr 28, 2017 at 11:14:24AM +0200, Thomas Monjalon wrote:
> 27/04/2017 11:25, Bruce Richardson:
> > On Thu, Apr 27, 2017 at 03:01:04AM -0400, Qi Zhang wrote:
> > > Vector PMD is not designed for i686 orginally, but it still can be active
> > > with i686 compile option.
> > > Below are observed failure when vPMD is invovled on i686
> > > (but may not limited to)
> > > 
> > > 1) memory overwrite when assign 2 mbuf points to rx return points.
> > > _mm_storeu_si128((__m128i *)&rx_pkts[pos+2], mbp2)
> > 
> > Is this a serious issue that prevents us using the driver? I think it's been
> > in the code for quite some time. Can it not be relatively easily fixed for
> > 32-bit builds?
> > 
> > > 
> > > 2) rearm_data is not 16 bytes aligned that cause general-protection exception
> > > _mm_store_si128((__m128i *)&rx_pkts[0]->rearm_data, rearm0);
> > > 
> > 
> > Good catch. I think this is also an easy fix. My preferred fix is to
> > explicitly align the rearm data on a 16-byte boundary. It would add some
> > padding to the middle of cacheline0 of the mbuf, but given that we
> > explicitly move other data to cacheline1, we will have padding on 32-bit
> > anyway, be it in the middle or the end of the mbuf cachelines.
> > 
> > > So the patch set will exclude Vector PMD from compile with i686 configure.
> 
> Please try to fix the drivers instead of turning them off.

The patch for alignment of physical addresses should fix these issues,
removing the need for these patches.

/Bruce

Re: [dpdk-dev] [PATCH 0/3 v2] disable vector PMD for i686

[Ferruh Yigit]

On 4/28/2017 10:26 AM, Bruce Richardson wrote:
> On Fri, Apr 28, 2017 at 11:14:24AM +0200, Thomas Monjalon wrote:
>> 27/04/2017 11:25, Bruce Richardson:
>>> On Thu, Apr 27, 2017 at 03:01:04AM -0400, Qi Zhang wrote:
>>>> Vector PMD is not designed for i686 orginally, but it still can be active
>>>> with i686 compile option.
>>>> Below are observed failure when vPMD is invovled on i686
>>>> (but may not limited to)
>>>>
>>>> 1) memory overwrite when assign 2 mbuf points to rx return points.
>>>> _mm_storeu_si128((__m128i *)&rx_pkts[pos+2], mbp2)
>>>
>>> Is this a serious issue that prevents us using the driver? I think it's been
>>> in the code for quite some time. Can it not be relatively easily fixed for
>>> 32-bit builds?
>>>
>>>>
>>>> 2) rearm_data is not 16 bytes aligned that cause general-protection exception
>>>> _mm_store_si128((__m128i *)&rx_pkts[0]->rearm_data, rearm0);
>>>>
>>>
>>> Good catch. I think this is also an easy fix. My preferred fix is to
>>> explicitly align the rearm data on a 16-byte boundary. It would add some
>>> padding to the middle of cacheline0 of the mbuf, but given that we
>>> explicitly move other data to cacheline1, we will have padding on 32-bit
>>> anyway, be it in the middle or the end of the mbuf cachelines.
>>>
>>>> So the patch set will exclude Vector PMD from compile with i686 configure.
>>
>> Please try to fix the drivers instead of turning them off.
> 
> The patch for alignment of physical addresses should fix these issues,
> removing the need for these patches.

Patches marked as rejected, Thanks for the fix Bruce!

> 
> /Bruce
>