Re: [dpdk-dev] [PATCH 1/4] eal: introduce the rte macro for always inline
On Sat, May 13, 2017 at 02:57:25PM +0530, Jerin Jacob wrote: > Different drivers use internal macros like force_inline for compiler > always inline feature. > Standardizing it through __rte_always_inline macro. > > Signed-off-by: Jerin Jacob > --- Good cleanup. Series Acked-by: Bruce Richardson
Re: [dpdk-dev] [PATCH 4/4] net/dpaa2: support parallel recv mode
On 5/12/2017 7:18 PM, Ferruh Yigit wrote:
On 4/19/2017 2:09 PM, Hemant Agrawal wrote:
Typically when the PMD issues a RX command to DPAA2 hardware,
the hw writes the available descriptors into the given memory.
The RX function then processes the frames and prepare them as
mbufs.
This patch adds support to issue another pull request to hardware
in another memory location, before we start processing the output
of the first request. This help in controlling the cpu cycles
wasted during the wait for the hardware to write the descriptors.
During hw debugging, it may be desired to keep the original
mode, so the original mode is also preserved and can be controlled
with an env flag.
Signed-off-by: Hemant Agrawal
<...>
+ /*If no prefetch is configured. */
+ if (getenv("DPAA2_RX_NO_PREFETCH")) {
Instead of getting configuration option from environment variable, can
you please make it argument to the driver?
This will be more consistent to the DPDK usage.
I guess you are talking about the devargs. That code is still not
cleaned up for non-pci devices.
In any case, we will work on that cleanup.
+ eth_dev->rx_pkt_burst = dpaa2_dev_rx;
+ PMD_INIT_LOG(INFO, "No Prefetch enabled");
+ }
+
<...>
Re: [dpdk-dev] [PATCH] drivers/net: document missing speed capabilities feature
On 5/12/2017 4:48 PM, Thomas Monjalon wrote: > 12/05/2017 12:49, Ferruh Yigit: >> On 5/10/2017 2:51 PM, Thomas Monjalon wrote: >>> 10/05/2017 15:10, Ferruh Yigit: --- a/doc/guides/nics/features/bnx2x.ini +++ b/doc/guides/nics/features/bnx2x.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = Y >>> >>> We should validate this feature only if the driver advertise the >>> right speeds for the device. >> >> Hi Thomas, >> >> Can you please clarify more, what is expected implementation in PMD? > > It is expected to advertise only the speeds that the device is > capable to offer. > >> And perhaps a good and a bad sample can be helpful. > > Good example: > drivers/net/i40e/i40e_ethdev.c > if (I40E_PHY_TYPE_SUPPORT_40G(hw->phy.phy_types)) > /* For XL710 */ > dev_info->speed_capa = ETH_LINK_SPEED_40G; > else if (I40E_PHY_TYPE_SUPPORT_25G(hw->phy.phy_types)) > /* For XXV710 */ > dev_info->speed_capa = ETH_LINK_SPEED_25G; > else > /* For X710 */ > dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G; > > Bad example: > drivers/net/bnx2x/bnx2x_ethdev.c > dev_info->speed_capa = ETH_LINK_SPEED_10G | ETH_LINK_SPEED_20G; > Looking at qlogic.com, only some 57840 adapters are capable of 20G. OK, I will update the patch according. Does it make sense to mark PMDs that report speed_capabilities, but without checking actual hw, as partial "P", to differentiate them from the ones that doesn't report at all, also this may help to PMD maintainers about what to fix. Thanks, ferruh
Re: [dpdk-dev] [PATCH] driver/net: remove unnecessary macro for unused variables
On 5/15/2017 3:31 AM, Yuanhan Liu wrote: > On Fri, May 12, 2017 at 11:33:03AM +0100, Ferruh Yigit wrote: >> remove __rte_unused instances that are not required. > > I'm wondering this is done by some scripts? Manually, unfortunately.. > > --yliu >
Re: [dpdk-dev] [PATCH] driver/net: remove unnecessary macro for unused variables
On 5/12/2017 11:33 AM, Ferruh Yigit wrote: > remove __rte_unused instances that are not required. > Hi Yuanhan, Does this kind of refactoring patches, specially the ones covering multiple drivers, cause trouble (more conflicts) to you while getting patches for stable trees? If so, I can postpone them through the end of integration deadline. But postponing them most probably will mean rebasing these patches later, so if this is not causing any problem for you, please let me know so that I can get them earlier. Thanks, ferruh
Re: [dpdk-dev] [PATCH] driver/net: remove unnecessary macro for unused variables
On Mon, May 15, 2017 at 10:17:43AM +0100, Ferruh Yigit wrote: > On 5/12/2017 11:33 AM, Ferruh Yigit wrote: > > remove __rte_unused instances that are not required. > > > > Hi Yuanhan, > > Does this kind of refactoring patches, specially the ones covering > multiple drivers, cause trouble (more conflicts) to you while getting > patches for stable trees? Yes, it's likely. > If so, I can postpone them through the end of integration deadline. I don't think it's necessary though. If a conflict happens, I will ask the author to do backport :) --yliu > But postponing them most probably will mean rebasing these patches > later, so if this is not causing any problem for you, please let me know > so that I can get them earlier. > > Thanks, > ferruh
Re: [dpdk-dev] [PATCH] driver/net: remove unnecessary macro for unused variables
On 5/15/2017 10:19 AM, Yuanhan Liu wrote: > On Mon, May 15, 2017 at 10:17:43AM +0100, Ferruh Yigit wrote: >> On 5/12/2017 11:33 AM, Ferruh Yigit wrote: >>> remove __rte_unused instances that are not required. >>> >> >> Hi Yuanhan, >> >> Does this kind of refactoring patches, specially the ones covering >> multiple drivers, cause trouble (more conflicts) to you while getting >> patches for stable trees? > > Yes, it's likely. > >> If so, I can postpone them through the end of integration deadline. > > I don't think it's necessary though. If a conflict happens, I will ask > the author to do backport :) OK, thanks for clarifying ... So, result is, refactoring patches won't be pushed through end of the release. Thanks, ferruh > > --yliu > >> But postponing them most probably will mean rebasing these patches >> later, so if this is not causing any problem for you, please let me know >> so that I can get them earlier. >> >> Thanks, >> ferruh
Re: [dpdk-dev] [PATCH] driver/net: remove unnecessary macro for unused variables
On Mon, May 15, 2017 at 10:26:00AM +0100, Ferruh Yigit wrote: > On 5/15/2017 10:19 AM, Yuanhan Liu wrote: > > On Mon, May 15, 2017 at 10:17:43AM +0100, Ferruh Yigit wrote: > >> On 5/12/2017 11:33 AM, Ferruh Yigit wrote: > >>> remove __rte_unused instances that are not required. > >>> > >> > >> Hi Yuanhan, > >> > >> Does this kind of refactoring patches, specially the ones covering > >> multiple drivers, cause trouble (more conflicts) to you while getting > >> patches for stable trees? > > > > Yes, it's likely. > > > >> If so, I can postpone them through the end of integration deadline. > > > > I don't think it's necessary though. If a conflict happens, I will ask > > the author to do backport :) > > OK, thanks for clarifying ... > > So, result is, refactoring patches won't be pushed through end of the > release. Yes, I see no strong need for that. --yliu
Re: [dpdk-dev] [PATCH] drivers/net: document missing speed capabilities feature
15/05/2017 10:40, Ferruh Yigit: > On 5/12/2017 4:48 PM, Thomas Monjalon wrote: > > 12/05/2017 12:49, Ferruh Yigit: > >> On 5/10/2017 2:51 PM, Thomas Monjalon wrote: > >>> 10/05/2017 15:10, Ferruh Yigit: > --- a/doc/guides/nics/features/bnx2x.ini > +++ b/doc/guides/nics/features/bnx2x.ini > @@ -4,6 +4,7 @@ > ; Refer to default.ini for the full list of available PMD features. > ; > [Features] > +Speed capabilities = Y > >>> > >>> We should validate this feature only if the driver advertise the > >>> right speeds for the device. > >> > >> Hi Thomas, > >> > >> Can you please clarify more, what is expected implementation in PMD? > > > > It is expected to advertise only the speeds that the device is > > capable to offer. > > > >> And perhaps a good and a bad sample can be helpful. > > > > Good example: > > drivers/net/i40e/i40e_ethdev.c > > if (I40E_PHY_TYPE_SUPPORT_40G(hw->phy.phy_types)) > > /* For XL710 */ > > dev_info->speed_capa = ETH_LINK_SPEED_40G; > > else if (I40E_PHY_TYPE_SUPPORT_25G(hw->phy.phy_types)) > > /* For XXV710 */ > > dev_info->speed_capa = ETH_LINK_SPEED_25G; > > else > > /* For X710 */ > > dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G; > > > > Bad example: > > drivers/net/bnx2x/bnx2x_ethdev.c > > dev_info->speed_capa = ETH_LINK_SPEED_10G | ETH_LINK_SPEED_20G; > > Looking at qlogic.com, only some 57840 adapters are capable of 20G. > > OK, I will update the patch according. > > Does it make sense to mark PMDs that report speed_capabilities, but > without checking actual hw, as partial "P", to differentiate them from > the ones that doesn't report at all, also this may help to PMD > maintainers about what to fix. Yes, we can use partial "P" here.
Re: [dpdk-dev] [PATCH v2] drivers/net/i40e/i40e_fdir.c: Improved i40e FDIR programming times
Hi Lijia,
> -Original Message-
> From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Michael Lilja
> Sent: Thursday, May 11, 2017 6:21 PM
> To: dev@dpdk.org
> Cc: Michael Lilja
> Subject: [dpdk-dev] [PATCH v2] drivers/net/i40e/i40e_fdir.c: Improved i40e
> FDIR programming times
>
> During my work (https://www.napatech.com/hw-acceleration-via-rte_flow/)
> on a flowtable application example that use rte_flow I discovered that the
> rte_flow programming times on a i40e was +11ms. The patch below result in
> an average programming time of 22usec with a max of 60usec instead of
> +11ms.
>
> Could the following patch be useful? There might be a good reason for the
> original code, I'm unable to tell, so I will let it up to the maintainer to
> decide.
Thanks for the patch, it's useful, and this can be removed from the commit log.
>
> Signed-off-by: Michael Lilja
>
> ---
> v2:
> * Code style fix
> ---
> drivers/net/i40e/i40e_fdir.c | 21 +
> 1 file changed, 13 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/net/i40e/i40e_fdir.c b/drivers/net/i40e/i40e_fdir.c index
> 28cc554f5..2162443f5 100644
> --- a/drivers/net/i40e/i40e_fdir.c
> +++ b/drivers/net/i40e/i40e_fdir.c
> @@ -1296,23 +1296,28 @@ i40e_fdir_filter_programming(struct i40e_pf *pf,
> rte_wmb();
> I40E_PCI_REG_WRITE(txq->qtx_tail, txq->tx_tail);
>
> - for (i = 0; i < I40E_FDIR_WAIT_COUNT; i++) {
> - rte_delay_us(I40E_FDIR_WAIT_INTERVAL_US);
> + for (i = 0; i < (I40E_FDIR_WAIT_COUNT *
> + I40E_FDIR_WAIT_INTERVAL_US); i++) {
> if ((txdp->cmd_type_offset_bsz &
> - rte_cpu_to_le_64(I40E_TXD_QW1_DTYPE_MASK)) ==
> -
> rte_cpu_to_le_64(I40E_TX_DESC_DTYPE_DESC_DONE))
> + rte_cpu_to_le_64(I40E_TXD_QW1_DTYPE_MASK)) ==
> + rte_cpu_to_le_64(I40E_TX_DESC_DTYPE_DESC_DONE))
> break;
> + rte_delay_us(1);
> }
> - if (i >= I40E_FDIR_WAIT_COUNT) {
> + if (i >= (I40E_FDIR_WAIT_COUNT * I40E_FDIR_WAIT_INTERVAL_US)) {
> PMD_DRV_LOG(ERR, "Failed to program FDIR filter:"
> - " time out to get DD on tx queue.");
> + " time out to get DD on tx queue.");
> return -ETIMEDOUT;
> }
> /* totally delay 10 ms to check programming status*/
> - rte_delay_us((I40E_FDIR_WAIT_COUNT - i) *
> I40E_FDIR_WAIT_INTERVAL_US);
> + for (i = 0; i < (I40E_FDIR_WAIT_COUNT *
> I40E_FDIR_WAIT_INTERVAL_US); i++) {
> + i
f (i40e_check_fdir_programming_status(rxq) >= 0) {
Braces {} can be removed here according to the coding style.
> + break;
How about "return 0;" here?
> + }
> + rte_delay_us(1);
> + }
> if (i40e_check_fdir_programming_status(rxq) < 0) {
How about removing the if statement? as i40e_check_fdir_programming_status(rxq)
has been executed in the above for loop.
> PMD_DRV_LOG(ERR, "Failed to program FDIR filter:"
> - " programming status reported.");
> + " programming status reported.");
> return -ENOSYS;
> }
>
> --
> 2.12.2
>
> Disclaimer: This email and any files transmitted with it may contain
> confidential information intended for the addressee(s) only. The information
> is not to be surrendered or copied to unauthorized persons. If you have
> received this communication in error, please notify the sender immediately
> and delete this e-mail from your system.
[dpdk-dev] [PATCH] cryptodev: remove crypto device type enumeration
Changes device type identification to be based on a unique
driver id replacing the current device type enumeration, which needed
library changes every time a new crypto driver was added.
The driver id is assigned dynamically during driver registration using
the new macro RTE_PMD_REGISTER_CRYPTO_DRIVER which returns a unique
uint8_t identifier for that driver. New APIs are also introduced
to allow retrieval of the driver id using the driver name.
Signed-off-by: Slawomir Mrozowicz
---
doc/guides/prog_guide/cryptodev_lib.rst| 5 +-
drivers/crypto/aesni_gcm/aesni_gcm_pmd.c | 12 +-
drivers/crypto/aesni_gcm/aesni_gcm_pmd_ops.c | 2 +-
drivers/crypto/aesni_mb/rte_aesni_mb_pmd.c | 12 +-
drivers/crypto/aesni_mb/rte_aesni_mb_pmd_ops.c | 2 +-
drivers/crypto/armv8/rte_armv8_pmd.c | 12 +-
drivers/crypto/armv8/rte_armv8_pmd_ops.c | 2 +-
drivers/crypto/dpaa2_sec/dpaa2_sec_dpseci.c| 10 +-
drivers/crypto/kasumi/rte_kasumi_pmd.c | 12 +-
drivers/crypto/kasumi/rte_kasumi_pmd_ops.c | 2 +-
drivers/crypto/null/null_crypto_pmd.c | 12 +-
drivers/crypto/null/null_crypto_pmd_ops.c | 2 +-
drivers/crypto/openssl/rte_openssl_pmd.c | 12 +-
drivers/crypto/openssl/rte_openssl_pmd_ops.c | 2 +-
drivers/crypto/qat/qat_crypto.c| 7 +-
drivers/crypto/qat/rte_qat_cryptodev.c | 8 +-
drivers/crypto/scheduler/rte_cryptodev_scheduler.c | 31 ++--
drivers/crypto/scheduler/scheduler_pmd.c | 7 +-
drivers/crypto/scheduler/scheduler_pmd_ops.c | 2 +-
drivers/crypto/scheduler/scheduler_pmd_private.h | 2 +-
drivers/crypto/snow3g/rte_snow3g_pmd.c | 12 +-
drivers/crypto/snow3g/rte_snow3g_pmd_ops.c | 2 +-
drivers/crypto/zuc/rte_zuc_pmd.c | 12 +-
drivers/crypto/zuc/rte_zuc_pmd_ops.c | 2 +-
lib/librte_cryptodev/rte_cryptodev.c | 39 -
lib/librte_cryptodev/rte_cryptodev.h | 68 ---
lib/librte_cryptodev/rte_cryptodev_pmd.h | 2 +-
lib/librte_cryptodev/rte_cryptodev_version.map | 11 +-
test/test/test_cryptodev.c | 195 ++---
test/test/test_cryptodev_blockcipher.c | 68 ---
test/test/test_cryptodev_blockcipher.h | 2 +-
test/test/test_cryptodev_perf.c| 124 -
32 files changed, 472 insertions(+), 221 deletions(-)
diff --git a/doc/guides/prog_guide/cryptodev_lib.rst
b/doc/guides/prog_guide/cryptodev_lib.rst
index 4f98f28..4644802 100644
--- a/doc/guides/prog_guide/cryptodev_lib.rst
+++ b/doc/guides/prog_guide/cryptodev_lib.rst
@@ -291,7 +291,7 @@ relevant information for the device.
struct rte_cryptodev_info {
const char *driver_name;
-enum rte_cryptodev_type dev_type;
+uint8_t driver_id;
struct rte_pci_device *pci_dev;
uint64_t feature_flags;
@@ -451,7 +451,8 @@ functions for the configuration of the session parameters
and freeing function
so the PMD can managed the memory on destruction of a session.
**Note**: Sessions created on a particular device can only be used on Crypto
-devices of the same type, and if you try to use a session on a device different
+devices of the same type - the same driver id used by this devices,
+and if you try to use a session on a device different
to that on which it was created then the Crypto operation will fail.
``rte_cryptodev_sym_session_create()`` is used to create a symmetric session on
diff --git a/drivers/crypto/aesni_gcm/aesni_gcm_pmd.c
b/drivers/crypto/aesni_gcm/aesni_gcm_pmd.c
index 101ef98..64a0ba0 100644
--- a/drivers/crypto/aesni_gcm/aesni_gcm_pmd.c
+++ b/drivers/crypto/aesni_gcm/aesni_gcm_pmd.c
@@ -143,8 +143,9 @@ aesni_gcm_get_session(struct aesni_gcm_qp *qp, struct
rte_crypto_sym_op *op)
struct aesni_gcm_session *sess = NULL;
if (op->sess_type == RTE_CRYPTO_SYM_OP_WITH_SESSION) {
- if (unlikely(op->session->dev_type
- != RTE_CRYPTODEV_AESNI_GCM_PMD))
+ if (unlikely(op->session->driver_id !=
+ rte_cryptodev_driver_id_get(
+ RTE_STR(CRYPTODEV_NAME_AESNI_GCM_PMD
return sess;
sess = (struct aesni_gcm_session *)op->session->_private;
@@ -456,7 +457,8 @@ aesni_gcm_create(const char *name,
goto init_error;
}
- dev->dev_type = RTE_CRYPTODEV_AESNI_GCM_PMD;
+ dev->driver_id = rte_cryptodev_driver_id_get(
+ RTE_STR(CRYPTODEV_NAME_AESNI_GCM_PMD));
dev->dev_ops = rte_aesni_gcm_pmd_ops;
/* register rx/tx burst functions for data path */
@@ -533,9 +535,13 @@ static struct rte_vdev_driver aesni_gcm_pmd_drv = {
.remove = aesni_gcm_remove
};
+static uint
Re: [dpdk-dev] [PATCH v1] doc: add template release notes for 17.08
11/05/2017 14:57, John McNamara: > Add template release notes for DPDK 17.08 with inline > comments and explanations of the various sections. > > Signed-off-by: John McNamara > --- > doc/guides/rel_notes/release_17_08.rst | 198 > + > 1 file changed, 198 insertions(+) > create mode 100644 doc/guides/rel_notes/release_17_08.rst Added to the index, and applied, thanks
[dpdk-dev] [PATCH] version: 17.08-rc0
Signed-off-by: Thomas Monjalon
---
lib/librte_eal/common/include/rte_version.h | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/lib/librte_eal/common/include/rte_version.h
b/lib/librte_eal/common/include/rte_version.h
index 07a085eb8..e9534fb25 100644
--- a/lib/librte_eal/common/include/rte_version.h
+++ b/lib/librte_eal/common/include/rte_version.h
@@ -61,7 +61,7 @@ extern "C" {
/**
* Minor version/month number i.e. the mm in yy.mm.z
*/
-#define RTE_VER_MONTH 5
+#define RTE_VER_MONTH 8
/**
* Patch level number i.e. the z in yy.mm.z
@@ -71,14 +71,14 @@ extern "C" {
/**
* Extra string to be appended to version number
*/
-#define RTE_VER_SUFFIX ""
+#define RTE_VER_SUFFIX "-rc"
/**
* Patch release number
* 0-15 = release candidates
* 16 = release
*/
-#define RTE_VER_RELEASE 16
+#define RTE_VER_RELEASE 0
/**
* Macro to compute a version number usable for comparisons
--
2.13.0
Re: [dpdk-dev] [PATCH v3 6/6] config: add clang support for armv8a linuxapp
On 5/12/2017 11:40 AM, Jianbo Liu wrote: On 12 May 2017 at 13:45, Ashwin Sekhar T K wrote: Moved all common defines from defconfig_arm64-armv8a-linuxapp-gcc to common_armv8a_linuxapp. Created new config arm64-armv8a-linuxapp-clang which adds the clang support to armv8a. Now defconfigs arm64-armv8a-linuxapp-gcc/clang contain only the CONFIG_RTE_TOOLCHAIN* defines and all other common defines are inherited from common_armv8a_linuxapp. Signed-off-by: Ashwin Sekhar T K Reviewed-by: Jerin Jacob --- config/common_armv8a_linuxapp| 53 config/defconfig_arm64-armv8a-linuxapp-clang | 35 ++ config/defconfig_arm64-armv8a-linuxapp-gcc | 23 +--- 3 files changed, 89 insertions(+), 22 deletions(-) create mode 100644 config/common_armv8a_linuxapp create mode 100644 config/defconfig_arm64-armv8a-linuxapp-clang Acked-by: Jianbo Liu Acked-by: Hemant Agrawal
Re: [dpdk-dev] [PATCH v3 0/6] Add clang compilation support for armv8a linuxapp
On 5/12/2017 11:15 AM, Ashwin Sekhar T K wrote: This series of patches adds the clang compilation support for armv8a linuxapp. Patch 1 adds an rte_pause() to a tight while loop in rte_eal_wait_lcore(). It fixes warning -Wempty-body seen with armv8a clang compilation. Patch 2 is basically for removing the usage of assembly directive ".arch armv8-a+crc" as this is not understood by clang. For removing these directives, compilation of armv8a crc32 support is made conditional and is only done for machines which has the crc extensions. Doing this avoids the need for having the ".arch armv8-a+crc" directives in the code. Patch 3, 4, and 5 are for fixing the compilation errors/warnings. Patch 6 adds the arm64-armv8a-linuxapp-clang defconfig. It also moves all the common defines to common_armv8a_config. Now defconfigs arm64-armv8a-linuxapp-gcc/clang contain only the defines related to toolchain. v3: * Moved [PATCH v2 6/6] to [PATCH v3 1/6] and moved [PATCH v2 2/6] to [PATCH v3 6/6]. v2: * Moved common defines from arm64-armv8a-linuxapp-gcc to common_armv8a_config * Removed the -Wno-empty-body flag from eal Makefile. Added rte_pause() to the while loop causing this warning. Ashwin Sekhar T K (6): eal: pause while busy-waiting for slave hash: compile armv8a CRC32 support conditionally net/thunderx: fix compile errors for armv8a clang acl: fix warning seen with armv8a clang eal/arm: fix warnings seen with armv8a clang config: add clang support for armv8a linuxapp config/common_armv8a_linuxapp | 53 ++ config/defconfig_arm64-armv8a-linuxapp-clang | 35 ++ config/defconfig_arm64-armv8a-linuxapp-gcc | 23 +- drivers/net/thunderx/base/nicvf_plat.h | 2 +- lib/librte_acl/Makefile| 5 +- lib/librte_eal/common/eal_common_launch.c | 3 +- .../common/include/arch/arm/rte_byteorder.h| 2 +- lib/librte_hash/Makefile | 2 + lib/librte_hash/rte_crc_arm64.h| 4 -- lib/librte_hash/rte_hash_crc.h | 2 +- 10 files changed, 100 insertions(+), 31 deletions(-) create mode 100644 config/common_armv8a_linuxapp create mode 100644 config/defconfig_arm64-armv8a-linuxapp-clang Series Acked-by: Hemant Agrawal
Re: [dpdk-dev] [PATCH] driver/net: remove unnecessary macro for unused variables
On 5/15/2017 1:53 AM, Lu, Wenzhuo wrote: > Hi, > >> -Original Message- >> From: Yigit, Ferruh >> Sent: Friday, May 12, 2017 6:33 PM >> To: John W. Linville; Legacy, Allain (Wind River); Peters, Matt (Wind River); >> Harish Patil; Rasesh Mody; Stephen Hurd; Ajit Khaparde; Doherty, Declan; Lu, >> Wenzhuo; Marcin Wojtas; Michal Krawczyk; Guy Tzalik; Evgeny Schemeilin; >> John Daley; Nelson Escobar; Chen, Jing D; Zhang, Helin; Wu, Jingjing; >> Ananyev, >> Konstantin; Andrew Rybchenko; Pascal Mazon; Yuanhan Liu; Maxime >> Coquelin; Shrikrishna Khare >> Cc: dev@dpdk.org; Yigit, Ferruh >> Subject: [PATCH] driver/net: remove unnecessary macro for unused variables >> >> remove __rte_unused instances that are not required. >> >> Signed-off-by: Ferruh Yigit Acked-by: Allain Legacy > Acked-by: Wenzhuo Lu Applied to dpdk-next-net/master, thanks.
[dpdk-dev] [PATCH v2] drivers/net: document missing speed capabilities feature
Signed-off-by: Ferruh Yigit --- Please fix partially support to fully supported speed capabilities. Please check other features of PMD and document missing ones in features file. Thanks. v2: * Mark PMDs that return speed capabilities without checking real hardware capabilities to Partially supported "P". * Mark PMDs return fake speed capabilities to "P", like virtio and vmxnet3. --- doc/guides/nics/features/ark.ini | 1 + doc/guides/nics/features/bnx2x.ini| 1 + doc/guides/nics/features/bnx2x_vf.ini | 1 + doc/guides/nics/features/cxgbe.ini| 1 + doc/guides/nics/features/dpaa2.ini| 1 + doc/guides/nics/features/e1000.ini| 1 + doc/guides/nics/features/ena.ini | 1 + doc/guides/nics/features/fm10k.ini| 1 + doc/guides/nics/features/fm10k_vec.ini| 1 + doc/guides/nics/features/fm10k_vf.ini | 1 + doc/guides/nics/features/fm10k_vf_vec.ini | 1 + doc/guides/nics/features/i40e.ini | 1 + doc/guides/nics/features/i40e_vec.ini | 1 + doc/guides/nics/features/igb.ini | 1 + doc/guides/nics/features/ixgbe.ini| 1 + doc/guides/nics/features/ixgbe_vec.ini| 1 + doc/guides/nics/features/mlx4.ini | 1 + doc/guides/nics/features/szedata2.ini | 1 + doc/guides/nics/features/tap.ini | 2 +- doc/guides/nics/features/virtio.ini | 1 + doc/guides/nics/features/virtio_vec.ini | 1 + doc/guides/nics/features/vmxnet3.ini | 1 + 22 files changed, 22 insertions(+), 1 deletion(-) diff --git a/doc/guides/nics/features/ark.ini b/doc/guides/nics/features/ark.ini index 31a3527..ec8a2b9 100644 --- a/doc/guides/nics/features/ark.ini +++ b/doc/guides/nics/features/ark.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = Y Queue start/stop = Y Jumbo frame = Y Scattered Rx = Y diff --git a/doc/guides/nics/features/bnx2x.ini b/doc/guides/nics/features/bnx2x.ini index 1ad8a3e..3e33e9a 100644 --- a/doc/guides/nics/features/bnx2x.ini +++ b/doc/guides/nics/features/bnx2x.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Link status event= Y Promiscuous mode = Y diff --git a/doc/guides/nics/features/bnx2x_vf.ini b/doc/guides/nics/features/bnx2x_vf.ini index da9168e..c270902 100644 --- a/doc/guides/nics/features/bnx2x_vf.ini +++ b/doc/guides/nics/features/bnx2x_vf.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Link status event= Y Promiscuous mode = Y diff --git a/doc/guides/nics/features/cxgbe.ini b/doc/guides/nics/features/cxgbe.ini index 2e72a10..2cf0af3 100644 --- a/doc/guides/nics/features/cxgbe.ini +++ b/doc/guides/nics/features/cxgbe.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Queue start/stop = Y MTU update = Y diff --git a/doc/guides/nics/features/dpaa2.ini b/doc/guides/nics/features/dpaa2.ini index d43f404..181f449 100644 --- a/doc/guides/nics/features/dpaa2.ini +++ b/doc/guides/nics/features/dpaa2.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Queue start/stop = Y MTU update = Y diff --git a/doc/guides/nics/features/e1000.ini b/doc/guides/nics/features/e1000.ini index 260d46d..51ca580 100644 --- a/doc/guides/nics/features/e1000.ini +++ b/doc/guides/nics/features/e1000.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Link status event= Y Rx interrupt = Y diff --git a/doc/guides/nics/features/ena.ini b/doc/guides/nics/features/ena.ini index 74969fd..691c1e3 100644 --- a/doc/guides/nics/features/ena.ini +++ b/doc/guides/nics/features/ena.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Queue start/stop = Y MTU update = Y Jumbo frame = Y diff --git a/doc/guides/nics/features/fm10k.ini b/doc/guides/nics/features/fm10k.ini index 9e1035f..f0f61a7 100644 --- a/doc/guides/nics/features/fm10k.ini +++ b/doc/guides/nics/features/fm10k.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Rx interrupt = Y Queue start/stop = Y Jumbo frame = Y diff --git a/doc/guides/nics/features/fm10k_vec.ini b/doc/guides/nics/features/fm10k_vec.ini index 1384ab1..4917e82 100644 --- a/doc/guides/nics/features/fm10k_vec.ini +++ b/doc/guides/nics/features/fm10k_vec.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full
Re: [dpdk-dev] [PATCH v2] drivers/net: document missing speed capabilities feature
Hi, In the case of tap PMD, I would say it's rather partially supported (P). Best regards, Pascal On Mon, 15 May 2017 11:59:46 +0100 Ferruh Yigit wrote: > Signed-off-by: Ferruh Yigit > --- > > Please fix partially support to fully supported speed capabilities. > > Please check other features of PMD and document missing ones in features > file. > > Thanks. > > v2: > * Mark PMDs that return speed capabilities without checking real > hardware capabilities to Partially supported "P". > * Mark PMDs return fake speed capabilities to "P", like virtio and > vmxnet3. > --- > doc/guides/nics/features/ark.ini | 1 + > doc/guides/nics/features/bnx2x.ini| 1 + > doc/guides/nics/features/bnx2x_vf.ini | 1 + > doc/guides/nics/features/cxgbe.ini| 1 + > doc/guides/nics/features/dpaa2.ini| 1 + > doc/guides/nics/features/e1000.ini| 1 + > doc/guides/nics/features/ena.ini | 1 + > doc/guides/nics/features/fm10k.ini| 1 + > doc/guides/nics/features/fm10k_vec.ini| 1 + > doc/guides/nics/features/fm10k_vf.ini | 1 + > doc/guides/nics/features/fm10k_vf_vec.ini | 1 + > doc/guides/nics/features/i40e.ini | 1 + > doc/guides/nics/features/i40e_vec.ini | 1 + > doc/guides/nics/features/igb.ini | 1 + > doc/guides/nics/features/ixgbe.ini| 1 + > doc/guides/nics/features/ixgbe_vec.ini| 1 + > doc/guides/nics/features/mlx4.ini | 1 + > doc/guides/nics/features/szedata2.ini | 1 + > doc/guides/nics/features/tap.ini | 2 +- > doc/guides/nics/features/virtio.ini | 1 + > doc/guides/nics/features/virtio_vec.ini | 1 + > doc/guides/nics/features/vmxnet3.ini | 1 + > 22 files changed, 22 insertions(+), 1 deletion(-) [...] > diff --git a/doc/guides/nics/features/tap.ini > b/doc/guides/nics/features/tap.ini > index 3efae75..d410409 100644 > --- a/doc/guides/nics/features/tap.ini > +++ b/doc/guides/nics/features/tap.ini > @@ -4,6 +4,7 @@ > ; Refer to default.ini for the full list of available PMD features. > ; > [Features] > +Speed capabilities = Y > Link status = Y > Link status event= Y > Jumbo frame = Y > @@ -13,7 +14,6 @@ Basic stats = Y > Flow API = Y > MTU update = Y > Multicast MAC filter = Y > -Speed capabilities = Y > Unicast MAC filter = Y > Packet type parsing = Y > Flow control = Y [...]
Re: [dpdk-dev] [PATCH v2] drivers/net: document missing speed capabilities feature
On 5/15/2017 1:17 PM, Pascal Mazon wrote: > Hi, > > In the case of tap PMD, I would say it's rather partially supported (P). Hi Pascal, Thanks, I will update it. > > Best regards, > > Pascal >
[dpdk-dev] [PATCH v3] drivers/net: document missing speed capabilities feature
Signed-off-by: Ferruh Yigit --- Please fix partially support to fully supported speed capabilities. Please check other features of PMD and document missing ones in features file. Thanks. v3: * Update tap PMD speed capability to "P" v2: * Mark PMDs that return speed capabilities without checking real hardware capabilities to Partially supported "P". * Mark PMDs return fake speed capabilities to "P", like virtio and vmxnet3. --- doc/guides/nics/features/ark.ini | 1 + doc/guides/nics/features/bnx2x.ini| 1 + doc/guides/nics/features/bnx2x_vf.ini | 1 + doc/guides/nics/features/cxgbe.ini| 1 + doc/guides/nics/features/dpaa2.ini| 1 + doc/guides/nics/features/e1000.ini| 1 + doc/guides/nics/features/ena.ini | 1 + doc/guides/nics/features/fm10k.ini| 1 + doc/guides/nics/features/fm10k_vec.ini| 1 + doc/guides/nics/features/fm10k_vf.ini | 1 + doc/guides/nics/features/fm10k_vf_vec.ini | 1 + doc/guides/nics/features/i40e.ini | 1 + doc/guides/nics/features/i40e_vec.ini | 1 + doc/guides/nics/features/igb.ini | 1 + doc/guides/nics/features/ixgbe.ini| 1 + doc/guides/nics/features/ixgbe_vec.ini| 1 + doc/guides/nics/features/mlx4.ini | 1 + doc/guides/nics/features/szedata2.ini | 1 + doc/guides/nics/features/tap.ini | 2 +- doc/guides/nics/features/virtio.ini | 1 + doc/guides/nics/features/virtio_vec.ini | 1 + doc/guides/nics/features/vmxnet3.ini | 1 + 22 files changed, 22 insertions(+), 1 deletion(-) diff --git a/doc/guides/nics/features/ark.ini b/doc/guides/nics/features/ark.ini index 31a3527..ec8a2b9 100644 --- a/doc/guides/nics/features/ark.ini +++ b/doc/guides/nics/features/ark.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = Y Queue start/stop = Y Jumbo frame = Y Scattered Rx = Y diff --git a/doc/guides/nics/features/bnx2x.ini b/doc/guides/nics/features/bnx2x.ini index 1ad8a3e..3e33e9a 100644 --- a/doc/guides/nics/features/bnx2x.ini +++ b/doc/guides/nics/features/bnx2x.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Link status event= Y Promiscuous mode = Y diff --git a/doc/guides/nics/features/bnx2x_vf.ini b/doc/guides/nics/features/bnx2x_vf.ini index da9168e..c270902 100644 --- a/doc/guides/nics/features/bnx2x_vf.ini +++ b/doc/guides/nics/features/bnx2x_vf.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Link status event= Y Promiscuous mode = Y diff --git a/doc/guides/nics/features/cxgbe.ini b/doc/guides/nics/features/cxgbe.ini index 2e72a10..2cf0af3 100644 --- a/doc/guides/nics/features/cxgbe.ini +++ b/doc/guides/nics/features/cxgbe.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Queue start/stop = Y MTU update = Y diff --git a/doc/guides/nics/features/dpaa2.ini b/doc/guides/nics/features/dpaa2.ini index d43f404..181f449 100644 --- a/doc/guides/nics/features/dpaa2.ini +++ b/doc/guides/nics/features/dpaa2.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Queue start/stop = Y MTU update = Y diff --git a/doc/guides/nics/features/e1000.ini b/doc/guides/nics/features/e1000.ini index 260d46d..51ca580 100644 --- a/doc/guides/nics/features/e1000.ini +++ b/doc/guides/nics/features/e1000.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Link status = Y Link status event= Y Rx interrupt = Y diff --git a/doc/guides/nics/features/ena.ini b/doc/guides/nics/features/ena.ini index 74969fd..691c1e3 100644 --- a/doc/guides/nics/features/ena.ini +++ b/doc/guides/nics/features/ena.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Queue start/stop = Y MTU update = Y Jumbo frame = Y diff --git a/doc/guides/nics/features/fm10k.ini b/doc/guides/nics/features/fm10k.ini index 9e1035f..f0f61a7 100644 --- a/doc/guides/nics/features/fm10k.ini +++ b/doc/guides/nics/features/fm10k.ini @@ -4,6 +4,7 @@ ; Refer to default.ini for the full list of available PMD features. ; [Features] +Speed capabilities = P Rx interrupt = Y Queue start/stop = Y Jumbo frame = Y diff --git a/doc/guides/nics/features/fm10k_vec.ini b/doc/guides/nics/features/fm10k_vec.ini index 1384ab1..4917e82 100644 --- a/doc/guides/nics/features/fm10k_vec.ini +++ b/doc/guides/nics/features/fm10k_vec.ini @@ -4,6
[dpdk-dev] SIMD Rx/Tx paths
Hi, I would like to open a discussion about SIMD code in drivers. I think we should not have different behaviours or features capabilities, in the different code paths of a same driver. I suggest to consider such differences as exceptions. So we should merge features files (i.e. matrix columns), and remove these files: % ls doc/guides/nics/features/*_vec.ini doc/guides/nics/features/fm10k_vec.ini doc/guides/nics/features/fm10k_vf_vec.ini doc/guides/nics/features/i40e_vec.ini doc/guides/nics/features/i40e_vf_vec.ini doc/guides/nics/features/ixgbe_vec.ini doc/guides/nics/features/ixgbe_vf_vec.ini doc/guides/nics/features/virtio_vec.ini If a feature is not supported in all code paths of a driver, it must be marked as partially (P) supported. Then the mid-term goal will be to try removing these inconsistencies. Opinions/comments?
[dpdk-dev] [PATCH v2 00/20] NXP DPAA2 PMD functional enhancements
This patchset adds following features to DPAA2 PMD: 1. improved error handling 2. parallel RX DQ requests to improve performance 3. Support for various ethdev operations 4. enable jumbo frames Hemant Agrawal (20): mk: change to cortex-a72 net/dpaa2: remove port level buffer layout definition net/dpaa2: stop using software annotation net/dpaa2: improve the error handling in dev init bus/fslmc: support for parallel Rx DQ requests net/dpaa2: support parallel Rx mode bus/fslmc: export qbman results in map file net/dpaa2: add support for congestion notification net/dpaa2: add support for tail drop on queue bus/fslmc: update TAILQ usages in dpaa2 objects net/dpaa2: add support for MAC address filtering net/dpaa2: add support for multicast promiscuous mode net/dpaa2: add support for VLAN filter and offload net/dpaa2: add support for VLAN strip net/dpaa2: add link status config support net/dpaa2: add support for flow control net/dpaa2: configure jumbo frames bus/fslmc: add support to detect soc version net/dpaa2: add support for Firmware Version get bus/fslmc: reducing the debug log messages config/defconfig_arm64-dpaa2-linuxapp-gcc | 2 +- doc/guides/nics/features/dpaa2.ini | 6 + drivers/bus/fslmc/Makefile | 1 + drivers/bus/fslmc/fslmc_vfio.c | 13 +- drivers/bus/fslmc/mc/dpmng.c| 81 drivers/bus/fslmc/mc/fsl_dpmng.h| 99 drivers/bus/fslmc/mc/fsl_dpmng_cmd.h| 54 +++ drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c| 30 +- drivers/bus/fslmc/portal/dpaa2_hw_dpio.c| 66 +-- drivers/bus/fslmc/portal/dpaa2_hw_pvt.h | 44 +- drivers/bus/fslmc/rte_bus_fslmc_version.map | 10 + drivers/net/dpaa2/base/dpaa2_hw_dpni.c | 19 +- drivers/net/dpaa2/dpaa2_ethdev.c| 693 +--- drivers/net/dpaa2/dpaa2_ethdev.h| 27 +- drivers/net/dpaa2/dpaa2_rxtx.c | 151 -- drivers/net/dpaa2/mc/dpni.c | 298 drivers/net/dpaa2/mc/fsl_dpni.h | 364 ++- drivers/net/dpaa2/mc/fsl_dpni_cmd.h | 142 ++ mk/machine/dpaa2/rte.vars.mk| 2 +- 19 files changed, 1915 insertions(+), 187 deletions(-) create mode 100644 drivers/bus/fslmc/mc/dpmng.c create mode 100644 drivers/bus/fslmc/mc/fsl_dpmng.h create mode 100644 drivers/bus/fslmc/mc/fsl_dpmng_cmd.h -- 1.9.1
[dpdk-dev] [PATCH v2 01/20] mk: change to cortex-a72
Signed-off-by: Hemant Agrawal --- config/defconfig_arm64-dpaa2-linuxapp-gcc | 2 +- mk/machine/dpaa2/rte.vars.mk | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/config/defconfig_arm64-dpaa2-linuxapp-gcc b/config/defconfig_arm64-dpaa2-linuxapp-gcc index 314a0ec..2304ab6 100644 --- a/config/defconfig_arm64-dpaa2-linuxapp-gcc +++ b/config/defconfig_arm64-dpaa2-linuxapp-gcc @@ -34,7 +34,7 @@ # NXP (Freescale) - Soc Architecture with WRIOP and QBMAN support CONFIG_RTE_MACHINE="dpaa2" -CONFIG_RTE_ARCH_ARM_TUNE="cortex-a57+fp+simd" +CONFIG_RTE_ARCH_ARM_TUNE="cortex-a72" # # Compile Environment Abstraction Layer diff --git a/mk/machine/dpaa2/rte.vars.mk b/mk/machine/dpaa2/rte.vars.mk index a60819f..675cd5a 100644 --- a/mk/machine/dpaa2/rte.vars.mk +++ b/mk/machine/dpaa2/rte.vars.mk @@ -57,5 +57,5 @@ MACHINE_CFLAGS += -march=armv8-a+crc ifdef CONFIG_RTE_ARCH_ARM_TUNE -MACHINE_CFLAGS += -mcpu=$(CONFIG_RTE_ARCH_ARM_TUNE:"%"=%) +MACHINE_CFLAGS += -mtune=$(CONFIG_RTE_ARCH_ARM_TUNE:"%"=%) endif -- 1.9.1
[dpdk-dev] [PATCH v2 03/20] net/dpaa2: stop using software annotation
The DPAA2 driver is not using the DPAA2 frame descriptor
software annotation area. This patch reduces the
PTA length to zero and adjust the RX Buffer Layout
accordingly.
Signed-off-by: Hemant Agrawal
---
drivers/bus/fslmc/portal/dpaa2_hw_pvt.h | 2 +-
drivers/net/dpaa2/base/dpaa2_hw_dpni.c | 19 ---
drivers/net/dpaa2/dpaa2_ethdev.c| 2 +-
3 files changed, 14 insertions(+), 9 deletions(-)
diff --git a/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
b/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
index c022373..9ffcec3 100644
--- a/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
+++ b/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
@@ -65,7 +65,7 @@
#define MAX_BPID 256
#define DPAA2_MBUF_HW_ANNOTATION 64
-#define DPAA2_FD_PTA_SIZE 64
+#define DPAA2_FD_PTA_SIZE 0
#if (DPAA2_MBUF_HW_ANNOTATION + DPAA2_FD_PTA_SIZE) > RTE_PKTMBUF_HEADROOM
#error "Annotation requirement is more than RTE_PKTMBUF_HEADROOM"
diff --git a/drivers/net/dpaa2/base/dpaa2_hw_dpni.c
b/drivers/net/dpaa2/base/dpaa2_hw_dpni.c
index 3dc60cc..547025d 100644
--- a/drivers/net/dpaa2/base/dpaa2_hw_dpni.c
+++ b/drivers/net/dpaa2/base/dpaa2_hw_dpni.c
@@ -306,15 +306,20 @@ int dpaa2_remove_flow_dist(
*/
/* ... rx buffer layout ... */
- tot_size = DPAA2_HW_BUF_RESERVE + RTE_PKTMBUF_HEADROOM;
- tot_size = RTE_ALIGN_CEIL(tot_size,
- DPAA2_PACKET_LAYOUT_ALIGN);
+ tot_size = RTE_PKTMBUF_HEADROOM;
+ tot_size = RTE_ALIGN_CEIL(tot_size, DPAA2_PACKET_LAYOUT_ALIGN);
memset(&layout, 0, sizeof(struct dpni_buffer_layout));
- layout.options = DPNI_BUF_LAYOUT_OPT_DATA_HEAD_ROOM;
-
- layout.data_head_room =
- tot_size - DPAA2_FD_PTA_SIZE - DPAA2_MBUF_HW_ANNOTATION;
+ layout.options = DPNI_BUF_LAYOUT_OPT_DATA_HEAD_ROOM |
+DPNI_BUF_LAYOUT_OPT_FRAME_STATUS |
+DPNI_BUF_LAYOUT_OPT_PARSER_RESULT |
+DPNI_BUF_LAYOUT_OPT_PRIVATE_DATA_SIZE;
+
+ layout.pass_frame_status = 1;
+ layout.private_data_size = DPAA2_FD_PTA_SIZE;
+ layout.pass_parser_result = 1;
+ layout.data_head_room = tot_size - DPAA2_FD_PTA_SIZE -
+ DPAA2_MBUF_HW_ANNOTATION;
retcode = dpni_set_buffer_layout(dpni, CMD_PRI_LOW, priv->token,
DPNI_QUEUE_RX, &layout);
if (retcode) {
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index f4c73de..007eda7 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -284,7 +284,7 @@
cfg.flc.value &= 0xFFC0;
/* 00 00 00 - last 6 bit represent annotation, context stashing,
* data stashing setting 01 01 00 (0x14) to enable
-* 1 line annotation, 1 line context
+* 1 line data, 1 line annotation
*/
cfg.flc.value |= 0x14;
}
--
1.9.1
[dpdk-dev] [PATCH v2 02/20] net/dpaa2: remove port level buffer layout definition
Signed-off-by: Hemant Agrawal
---
drivers/net/dpaa2/dpaa2_ethdev.c | 25 -
1 file changed, 25 deletions(-)
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 4576442..f4c73de 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -761,7 +761,6 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
struct dpaa2_dev_priv *priv = eth_dev->data->dev_private;
struct dpni_buffer_layout layout;
int i, ret, hw_id;
- int tot_size;
PMD_INIT_FUNC_TRACE();
@@ -854,30 +853,6 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
return -ret;
}
- /* ... rx buffer layout ... */
- tot_size = DPAA2_HW_BUF_RESERVE + RTE_PKTMBUF_HEADROOM;
- tot_size = RTE_ALIGN_CEIL(tot_size,
- DPAA2_PACKET_LAYOUT_ALIGN);
-
- memset(&layout, 0, sizeof(struct dpni_buffer_layout));
- layout.options = DPNI_BUF_LAYOUT_OPT_FRAME_STATUS |
- DPNI_BUF_LAYOUT_OPT_PARSER_RESULT |
- DPNI_BUF_LAYOUT_OPT_DATA_HEAD_ROOM |
- DPNI_BUF_LAYOUT_OPT_PRIVATE_DATA_SIZE;
-
- layout.pass_frame_status = 1;
- layout.data_head_room = tot_size
- - DPAA2_FD_PTA_SIZE - DPAA2_MBUF_HW_ANNOTATION;
- layout.private_data_size = DPAA2_FD_PTA_SIZE;
- layout.pass_parser_result = 1;
- PMD_INIT_LOG(DEBUG, "Tot_size = %d, head room = %d, private = %d",
-tot_size, layout.data_head_room, layout.private_data_size);
- ret = dpni_set_buffer_layout(dpni_dev, CMD_PRI_LOW, priv->token,
-DPNI_QUEUE_RX, &layout);
- if (ret) {
- PMD_INIT_LOG(ERR, "Err(%d) in setting rx buffer layout", ret);
- return -1;
- }
/* ... tx buffer layout ... */
memset(&layout, 0, sizeof(struct dpni_buffer_layout));
--
1.9.1
[dpdk-dev] [PATCH v2 04/20] net/dpaa2: improve the error handling in dev init
Signed-off-by: Hemant Agrawal
---
drivers/net/dpaa2/dpaa2_ethdev.c | 69 +++-
1 file changed, 39 insertions(+), 30 deletions(-)
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 007eda7..698fb84 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -54,6 +54,7 @@
#include "dpaa2_ethdev.h"
static struct rte_dpaa2_driver rte_dpaa2_pmd;
+static int dpaa2_dev_uninit(struct rte_eth_dev *eth_dev);
/**
* Atomically reads the link status information from global
@@ -772,7 +773,7 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
hw_id = dpaa2_dev->object_id;
- dpni_dev = (struct fsl_mc_io *)malloc(sizeof(struct fsl_mc_io));
+ dpni_dev = rte_malloc(NULL, sizeof(struct fsl_mc_io), 0);
if (!dpni_dev) {
PMD_INIT_LOG(ERR, "malloc failed for dpni device\n");
return -1;
@@ -781,24 +782,28 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
dpni_dev->regs = rte_mcp_ptr_list[0];
ret = dpni_open(dpni_dev, CMD_PRI_LOW, hw_id, &priv->token);
if (ret) {
- PMD_INIT_LOG(ERR, "Failure in opening dpni@%d device with"
- " error code %d\n", hw_id, ret);
+ PMD_INIT_LOG(ERR,
+"Failure in opening dpni@%d with err code %d\n",
+hw_id, ret);
+ rte_free(dpni_dev);
return -1;
}
/* Clean the device first */
ret = dpni_reset(dpni_dev, CMD_PRI_LOW, priv->token);
if (ret) {
- PMD_INIT_LOG(ERR, "Failure cleaning dpni@%d device with"
- " error code %d\n", hw_id, ret);
- return -1;
+ PMD_INIT_LOG(ERR,
+"Failure cleaning dpni@%d with err code %d\n",
+hw_id, ret);
+ goto init_err;
}
ret = dpni_get_attributes(dpni_dev, CMD_PRI_LOW, priv->token, &attr);
if (ret) {
- PMD_INIT_LOG(ERR, "Failure in getting dpni@%d attribute, "
- " error code %d\n", hw_id, ret);
- return -1;
+ PMD_INIT_LOG(ERR,
+"Failure in get dpni@%d attribute, err code %d\n",
+hw_id, ret);
+ goto init_err;
}
priv->num_tc = attr.num_tcs;
@@ -831,29 +836,29 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
ret = dpaa2_alloc_rx_tx_queues(eth_dev);
if (ret) {
PMD_INIT_LOG(ERR, "dpaa2_alloc_rx_tx_queuesFailed\n");
- return -ret;
+ goto init_err;
}
/* Allocate memory for storing MAC addresses */
eth_dev->data->mac_addrs = rte_zmalloc("dpni",
ETHER_ADDR_LEN * attr.mac_filter_entries, 0);
if (eth_dev->data->mac_addrs == NULL) {
- PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
- "store MAC addresses",
- ETHER_ADDR_LEN * attr.mac_filter_entries);
- return -ENOMEM;
+ PMD_INIT_LOG(ERR,
+ "Failed to allocate %d bytes needed to store MAC addresses",
+ETHER_ADDR_LEN * attr.mac_filter_entries);
+ ret = -ENOMEM;
+ goto init_err;
}
ret = dpni_get_primary_mac_addr(dpni_dev, CMD_PRI_LOW,
priv->token,
(uint8_t *)(eth_dev->data->mac_addrs[0].addr_bytes));
if (ret) {
- PMD_INIT_LOG(ERR, "DPNI get mac address failed:"
- " Error Code = %d\n", ret);
- return -ret;
+ PMD_INIT_LOG(ERR, "DPNI get mac address failed:Err Code = %d\n",
+ret);
+ goto init_err;
}
-
/* ... tx buffer layout ... */
memset(&layout, 0, sizeof(struct dpni_buffer_layout));
layout.options = DPNI_BUF_LAYOUT_OPT_FRAME_STATUS;
@@ -861,9 +866,9 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
ret = dpni_set_buffer_layout(dpni_dev, CMD_PRI_LOW, priv->token,
DPNI_QUEUE_TX, &layout);
if (ret) {
- PMD_INIT_LOG(ERR, "Error (%d) in setting tx buffer"
- " layout", ret);
- return -1;
+ PMD_INIT_LOG(ERR, "Error (%d) in setting tx buffer layout",
+ret);
+ goto init_err;
}
/* ... tx-conf and error buffer layout ... */
@@ -873,9 +878,9 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
ret = dpni_set_buffer_layout(dpni_dev, CMD_PRI_LOW, priv->token,
DPNI_QUEUE_TX_CONFIRM, &layout);
[dpdk-dev] [PATCH v2 05/20] bus/fslmc: support for parallel Rx DQ requests
DPAA2 hardware support the option to configure
multiple memories for Rx recv buffer (DPAA2-DQRR).
Each Rx request executing is called as 'DQ' request.
This patch adds routines to get information w.r.t each DQ request.
Signed-off-by: Hemant Agrawal
---
drivers/bus/fslmc/portal/dpaa2_hw_dpio.c | 2 ++
drivers/bus/fslmc/portal/dpaa2_hw_pvt.h | 37
2 files changed, 39 insertions(+)
diff --git a/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
b/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
index a1a58b9..203a7d5 100644
--- a/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
+++ b/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
@@ -67,6 +67,8 @@
struct dpaa2_io_portal_t dpaa2_io_portal[RTE_MAX_LCORE];
RTE_DEFINE_PER_LCORE(struct dpaa2_io_portal_t, _dpaa2_io);
+struct swp_active_dqs global_active_dqs_list[NUM_MAX_SWP];
+
TAILQ_HEAD(dpio_device_list, dpaa2_dpio_dev);
static struct dpio_device_list *dpio_dev_list; /*!< DPIO device list */
static uint32_t io_space_count;
diff --git a/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
b/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
index 9ffcec3..09e9398 100644
--- a/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
+++ b/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
@@ -108,6 +108,9 @@ struct dpaa2_dpbp_dev {
struct queue_storage_info_t {
struct qbman_result *dq_storage[NUM_DQS_PER_QUEUE];
+ struct qbman_result *active_dqs;
+ int active_dpio_id;
+ int toggle;
};
struct dpaa2_queue {
@@ -123,6 +126,15 @@ struct dpaa2_queue {
struct queue_storage_info_t *q_storage;
};
+struct swp_active_dqs {
+ struct qbman_result *global_active_dqs;
+ uint64_t reserved[7];
+};
+
+#define NUM_MAX_SWP 64
+
+extern struct swp_active_dqs global_active_dqs_list[NUM_MAX_SWP];
+
/*! Global MCP list */
extern void *(*rte_mcp_ptr_list);
@@ -264,6 +276,31 @@ static phys_addr_t dpaa2_mem_vtop(uint64_t vaddr)
#endif /* RTE_LIBRTE_DPAA2_USE_PHYS_IOVA */
+static inline
+int check_swp_active_dqs(uint16_t dpio_dev_index)
+{
+ if (global_active_dqs_list[dpio_dev_index].global_active_dqs != NULL)
+ return 1;
+ return 0;
+}
+
+static inline
+void clear_swp_active_dqs(uint16_t dpio_dev_index)
+{
+ global_active_dqs_list[dpio_dev_index].global_active_dqs = NULL;
+}
+
+static inline
+struct qbman_result *get_swp_active_dqs(uint16_t dpio_dev_index)
+{
+ return global_active_dqs_list[dpio_dev_index].global_active_dqs;
+}
+
+static inline
+void set_swp_active_dqs(uint16_t dpio_dev_index, struct qbman_result *dqs)
+{
+ global_active_dqs_list[dpio_dev_index].global_active_dqs = dqs;
+}
struct dpaa2_dpbp_dev *dpaa2_alloc_dpbp_dev(void);
void dpaa2_free_dpbp_dev(struct dpaa2_dpbp_dev *dpbp);
--
1.9.1
[dpdk-dev] [PATCH v2 07/20] bus/fslmc: export qbman results in map file
Signed-off-by: Hemant Agrawal
---
drivers/bus/fslmc/rte_bus_fslmc_version.map | 8
1 file changed, 8 insertions(+)
diff --git a/drivers/bus/fslmc/rte_bus_fslmc_version.map
b/drivers/bus/fslmc/rte_bus_fslmc_version.map
index 2db0fce..6e7b94d 100644
--- a/drivers/bus/fslmc/rte_bus_fslmc_version.map
+++ b/drivers/bus/fslmc/rte_bus_fslmc_version.map
@@ -49,3 +49,11 @@ DPDK_17.05 {
local: *;
};
+
+DPDK_17.08 {
+ global:
+
+ qbman_result_SCN_state_in_mem;
+
+ local: *;
+};
--
1.9.1
[dpdk-dev] [PATCH v2 08/20] net/dpaa2: add support for congestion notification
Signed-off-by: Hemant Agrawal
---
drivers/bus/fslmc/portal/dpaa2_hw_pvt.h | 5 +-
drivers/net/dpaa2/dpaa2_ethdev.c| 62 ++-
drivers/net/dpaa2/dpaa2_ethdev.h| 14
drivers/net/dpaa2/dpaa2_rxtx.c | 4 +
drivers/net/dpaa2/mc/dpni.c | 47
drivers/net/dpaa2/mc/fsl_dpni.h | 129 +++-
drivers/net/dpaa2/mc/fsl_dpni_cmd.h | 36 +
7 files changed, 289 insertions(+), 8 deletions(-)
diff --git a/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
b/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
index 09e9398..6549b84 100644
--- a/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
+++ b/drivers/bus/fslmc/portal/dpaa2_hw_pvt.h
@@ -123,7 +123,10 @@ struct dpaa2_queue {
uint64_t rx_pkts;
uint64_t tx_pkts;
uint64_t err_pkts;
- struct queue_storage_info_t *q_storage;
+ union {
+ struct queue_storage_info_t *q_storage;
+ struct qbman_result *cscn;
+ };
};
struct swp_active_dqs {
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 797e71d..80f1cd7 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -177,8 +177,13 @@
for (i = 0; i < priv->nb_tx_queues; i++) {
mc_q->dev = dev;
- mc_q->flow_id = DPNI_NEW_FLOW_ID;
+ mc_q->flow_id = 0x;
priv->tx_vq[i] = mc_q++;
+ dpaa2_q = (struct dpaa2_queue *)priv->tx_vq[i];
+ dpaa2_q->cscn = rte_malloc(NULL,
+ sizeof(struct qbman_result), 16);
+ if (!dpaa2_q->cscn)
+ goto fail_tx;
}
vq_id = 0;
@@ -191,6 +196,14 @@
}
return 0;
+fail_tx:
+ i -= 1;
+ while (i >= 0) {
+ dpaa2_q = (struct dpaa2_queue *)priv->tx_vq[i];
+ rte_free(dpaa2_q->cscn);
+ priv->tx_vq[i--] = NULL;
+ }
+ i = priv->nb_rx_queues;
fail:
i -= 1;
mc_q = priv->rx_vq[0];
@@ -320,7 +333,7 @@
PMD_INIT_FUNC_TRACE();
/* Return if queue already configured */
- if (dpaa2_q->flow_id != DPNI_NEW_FLOW_ID)
+ if (dpaa2_q->flow_id != 0x)
return 0;
memset(&tx_conf_cfg, 0, sizeof(struct dpni_queue));
@@ -358,6 +371,36 @@
}
dpaa2_q->tc_index = tc_id;
+ if (priv->flags & DPAA2_TX_CGR_SUPPORT) {
+ struct dpni_congestion_notification_cfg cong_notif_cfg;
+
+ cong_notif_cfg.units = DPNI_CONGESTION_UNIT_BYTES;
+ /* Notify about congestion when the queue size is 32 KB */
+ cong_notif_cfg.threshold_entry = CONG_ENTER_TX_THRESHOLD;
+ /* Notify that the queue is not congested when the data in
+* the queue is below this thershold.
+*/
+ cong_notif_cfg.threshold_exit = CONG_EXIT_TX_THRESHOLD;
+ cong_notif_cfg.message_ctx = 0;
+ cong_notif_cfg.message_iova = (uint64_t)dpaa2_q->cscn;
+ cong_notif_cfg.dest_cfg.dest_type = DPNI_DEST_NONE;
+ cong_notif_cfg.notification_mode =
+DPNI_CONG_OPT_WRITE_MEM_ON_ENTER |
+DPNI_CONG_OPT_WRITE_MEM_ON_EXIT |
+DPNI_CONG_OPT_COHERENT_WRITE;
+
+ ret = dpni_set_congestion_notification(dpni, CMD_PRI_LOW,
+ priv->token,
+ DPNI_QUEUE_TX,
+ tc_id,
+ &cong_notif_cfg);
+ if (ret) {
+ PMD_INIT_LOG(ERR,
+ "Error in setting tx congestion notification: = %d",
+ -ret);
+ return -ret;
+ }
+ }
dev->data->tx_queues[tx_queue_id] = dpaa2_q;
return 0;
}
@@ -513,12 +556,22 @@
static void
dpaa2_dev_close(struct rte_eth_dev *dev)
{
+ struct rte_eth_dev_data *data = dev->data;
struct dpaa2_dev_priv *priv = dev->data->dev_private;
struct fsl_mc_io *dpni = (struct fsl_mc_io *)priv->hw;
- int ret;
+ int i, ret;
+ struct dpaa2_queue *dpaa2_q;
PMD_INIT_FUNC_TRACE();
+ for (i = 0; i < data->nb_tx_queues; i++) {
+ dpaa2_q = (struct dpaa2_queue *)data->tx_queues[i];
+ if (!dpaa2_q->cscn) {
+ rte_free(dpaa2_q->cscn);
+ dpaa2_q->cscn = NULL;
+ }
+ }
+
/* Clean the device first */
ret = dpni_reset(dpni, CMD_PRI_LOW, priv->token);
if (ret) {
@@ -832,6 +885,9 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
priv->max
[dpdk-dev] [PATCH v2 06/20] net/dpaa2: support parallel Rx mode
Typically when the PMD issues a RX command to DPAA2 hardware,
the hw writes the available descriptors into the given memory.
The RX function then processes the frames and prepare them as
mbufs.
This patch adds support to issue another pull request to hardware
in another memory location, before we start processing the output
of the first request. This help in controlling the cpu cycles
wasted during the wait for the hardware to write the descriptors.
Signed-off-by: Hemant Agrawal
---
drivers/net/dpaa2/dpaa2_ethdev.c | 4 +-
drivers/net/dpaa2/dpaa2_ethdev.h | 3 +-
drivers/net/dpaa2/dpaa2_rxtx.c | 119 ---
3 files changed, 79 insertions(+), 47 deletions(-)
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 698fb84..797e71d 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -391,7 +391,7 @@
RTE_PTYPE_UNKNOWN
};
- if (dev->rx_pkt_burst == dpaa2_dev_rx)
+ if (dev->rx_pkt_burst == dpaa2_dev_prefetch_rx)
return ptypes;
return NULL;
}
@@ -886,7 +886,7 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
eth_dev->dev_ops = &dpaa2_ethdev_ops;
eth_dev->data->drv_name = rte_dpaa2_pmd.driver.name;
- eth_dev->rx_pkt_burst = dpaa2_dev_rx;
+ eth_dev->rx_pkt_burst = dpaa2_dev_prefetch_rx;
eth_dev->tx_pkt_burst = dpaa2_dev_tx;
rte_fslmc_vfio_dmamap();
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.h b/drivers/net/dpaa2/dpaa2_ethdev.h
index 7196398..7fa7e7d 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.h
+++ b/drivers/net/dpaa2/dpaa2_ethdev.h
@@ -77,7 +77,8 @@ int dpaa2_remove_flow_dist(struct rte_eth_dev *eth_dev,
int dpaa2_attach_bp_list(struct dpaa2_dev_priv *priv, void *blist);
-uint16_t dpaa2_dev_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts);
+uint16_t dpaa2_dev_prefetch_rx(void *queue, struct rte_mbuf **bufs,
+ uint16_t nb_pkts);
uint16_t dpaa2_dev_tx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts);
#endif /* _DPAA2_ETHDEV_H */
diff --git a/drivers/net/dpaa2/dpaa2_rxtx.c b/drivers/net/dpaa2/dpaa2_rxtx.c
index c5d49cb..9b7539a 100644
--- a/drivers/net/dpaa2/dpaa2_rxtx.c
+++ b/drivers/net/dpaa2/dpaa2_rxtx.c
@@ -242,17 +242,18 @@ static inline int __attribute__((hot))
}
uint16_t
-dpaa2_dev_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
+dpaa2_dev_prefetch_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
{
- /* Function is responsible to receive frames for a given device and VQ*/
+ /* Function receive frames for a given device and VQ*/
struct dpaa2_queue *dpaa2_q = (struct dpaa2_queue *)queue;
struct qbman_result *dq_storage;
uint32_t fqid = dpaa2_q->fqid;
int ret, num_rx = 0;
uint8_t is_last = 0, status;
struct qbman_swp *swp;
- const struct qbman_fd *fd;
+ const struct qbman_fd *fd[DPAA2_DQRR_RING_SIZE];
struct qbman_pull_desc pulldesc;
+ struct queue_storage_info_t *q_storage = dpaa2_q->q_storage;
struct rte_eth_dev *dev = dpaa2_q->dev;
if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
@@ -263,44 +264,51 @@ static inline int __attribute__((hot))
}
}
swp = DPAA2_PER_LCORE_PORTAL;
- dq_storage = dpaa2_q->q_storage->dq_storage[0];
-
- qbman_pull_desc_clear(&pulldesc);
- qbman_pull_desc_set_numframes(&pulldesc,
- (nb_pkts > DPAA2_DQRR_RING_SIZE) ?
- DPAA2_DQRR_RING_SIZE : nb_pkts);
- qbman_pull_desc_set_fq(&pulldesc, fqid);
- /* todo optimization - we can have dq_storage_phys available*/
- qbman_pull_desc_set_storage(&pulldesc, dq_storage,
+ if (!q_storage->active_dqs) {
+ q_storage->toggle = 0;
+ dq_storage = q_storage->dq_storage[q_storage->toggle];
+ qbman_pull_desc_clear(&pulldesc);
+ qbman_pull_desc_set_numframes(&pulldesc,
+ (nb_pkts > DPAA2_DQRR_RING_SIZE) ?
+ DPAA2_DQRR_RING_SIZE : nb_pkts);
+ qbman_pull_desc_set_fq(&pulldesc, fqid);
+ qbman_pull_desc_set_storage(&pulldesc, dq_storage,
(dma_addr_t)(DPAA2_VADDR_TO_IOVA(dq_storage)), 1);
-
- /*Issue a volatile dequeue command. */
- while (1) {
- if (qbman_swp_pull(swp, &pulldesc)) {
- PMD_RX_LOG(ERR, "VDQ command is not issued."
- "QBMAN is busy\n");
- /* Portal was busy, try again */
- continue;
+ if (check_swp_active_dqs(DPAA2_PER_LCORE_DPIO->index)) {
+ while (!qbman_check_command_complete(swp,
+ get_swp_active_dqs(DPAA2_PER_LCORE_DPIO->index)))
[dpdk-dev] [PATCH v2 09/20] net/dpaa2: add support for tail drop on queue
This will help in limiting the size of queues and avoid
them growing practicaly infinite.
Signed-off-by: Hemant Agrawal
---
drivers/net/dpaa2/dpaa2_ethdev.c| 19 +
drivers/net/dpaa2/dpaa2_ethdev.h| 8
drivers/net/dpaa2/mc/dpni.c | 50 ++
drivers/net/dpaa2/mc/fsl_dpni.h | 85 +
drivers/net/dpaa2/mc/fsl_dpni_cmd.h | 29 +
5 files changed, 191 insertions(+)
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 80f1cd7..0526e26 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -309,6 +309,25 @@
return -1;
}
+ if (!(priv->flags & DPAA2_RX_TAILDROP_OFF)) {
+ struct dpni_taildrop taildrop;
+
+ taildrop.enable = 1;
+ /*enabling per rx queue congestion control */
+ taildrop.threshold = CONG_THRESHOLD_RX_Q;
+ taildrop.units = DPNI_CONGESTION_UNIT_BYTES;
+ PMD_INIT_LOG(DEBUG, "Enabling Early Drop on queue = %d",
+rx_queue_id);
+ ret = dpni_set_taildrop(dpni, CMD_PRI_LOW, priv->token,
+ DPNI_CP_QUEUE, DPNI_QUEUE_RX,
+ dpaa2_q->tc_index, flow_id, &taildrop);
+ if (ret) {
+ PMD_INIT_LOG(ERR, "Error in setting the rx flow"
+" err : = %d\n", ret);
+ return -1;
+ }
+ }
+
dev->data->rx_queues[rx_queue_id] = dpaa2_q;
return 0;
}
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.h b/drivers/net/dpaa2/dpaa2_ethdev.h
index e7728ba..3254b99 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.h
+++ b/drivers/net/dpaa2/dpaa2_ethdev.h
@@ -56,6 +56,11 @@
*/
#define CONG_EXIT_TX_THRESHOLD(24 * 1024)
+/* RX queue tail drop threshold
+ * currently considering 32 KB packets
+ */
+#define CONG_THRESHOLD_RX_Q (32 * 1024)
+
/* Size of the input SMMU mapped memory required by MC */
#define DIST_PARAM_IOVA_SIZE 256
@@ -64,6 +69,9 @@
*/
#define DPAA2_TX_CGR_SUPPORT 0x01
+/* Disable RX tail drop, default is enable */
+#define DPAA2_RX_TAILDROP_OFF 0x04
+
struct dpaa2_dev_priv {
void *hw;
int32_t hw_id;
diff --git a/drivers/net/dpaa2/mc/dpni.c b/drivers/net/dpaa2/mc/dpni.c
index 145336d..ad33282 100644
--- a/drivers/net/dpaa2/mc/dpni.c
+++ b/drivers/net/dpaa2/mc/dpni.c
@@ -784,3 +784,53 @@ int dpni_reset_statistics(struct fsl_mc_io *mc_io,
/* send command to mc*/
return mc_send_command(mc_io, &cmd);
}
+
+int dpni_set_taildrop(struct fsl_mc_io *mc_io,
+ uint32_t cmd_flags,
+uint16_t token,
+enum dpni_congestion_point cg_point,
+enum dpni_queue_type q_type,
+uint8_t tc,
+uint8_t q_index,
+struct dpni_taildrop *taildrop)
+{
+ struct mc_command cmd = { 0 };
+
+ /* prepare command */
+ cmd.header = mc_encode_cmd_header(DPNI_CMDID_SET_TAILDROP,
+ cmd_flags,
+ token);
+ DPNI_CMD_SET_TAILDROP(cmd, cg_point, q_type, tc, q_index, taildrop);
+
+ /* send command to mc*/
+ return mc_send_command(mc_io, &cmd);
+}
+
+int dpni_get_taildrop(struct fsl_mc_io *mc_io,
+ uint32_t cmd_flags,
+uint16_t token,
+enum dpni_congestion_point cg_point,
+enum dpni_queue_type q_type,
+uint8_t tc,
+uint8_t q_index,
+struct dpni_taildrop *taildrop)
+{
+ struct mc_command cmd = { 0 };
+ int err;
+
+ /* prepare command */
+ cmd.header = mc_encode_cmd_header(DPNI_CMDID_GET_TAILDROP,
+ cmd_flags,
+ token);
+ DPNI_CMD_GET_TAILDROP(cmd, cg_point, q_type, tc, q_index);
+
+ /* send command to mc*/
+ err = mc_send_command(mc_io, &cmd);
+ if (err)
+ return err;
+
+ /* retrieve response parameters */
+ DPNI_RSP_GET_TAILDROP(cmd, taildrop);
+
+ return 0;
+}
diff --git a/drivers/net/dpaa2/mc/fsl_dpni.h b/drivers/net/dpaa2/mc/fsl_dpni.h
index 10dccc7..68e30df 100644
--- a/drivers/net/dpaa2/mc/fsl_dpni.h
+++ b/drivers/net/dpaa2/mc/fsl_dpni.h
@@ -1335,4 +1335,89 @@ int dpni_reset_statistics(struct fsl_mc_io *mc_io,
uint32_t cmd_flags,
uint16_t token);
+/**
+ * enum dpni_congestion_point - Structure representing congestion point
+ * @DPNI_CP_QUEUE: Set taildrop per queue, identified by QUEUE_TYPE, TC and
+ * QUEUE_INDEX
+ * @DPNI_CP_GROUP: Set tail
[dpdk-dev] [PATCH v2 06/20] net/dpaa2: support parallel Rx mode
Typically when the PMD issues a RX command to DPAA2 hardware,
the hw writes the available descriptors into the given memory.
The RX function then processes the frames and prepare them as
mbufs.
This patch adds support to issue another pull request to hardware
in another memory location, before we start processing the output
of the first request. This help in controlling the cpu cycles
wasted during the wait for the hardware to write the descriptors.
Signed-off-by: Hemant Agrawal
---
drivers/net/dpaa2/dpaa2_ethdev.c | 4 +-
drivers/net/dpaa2/dpaa2_ethdev.h | 3 +-
drivers/net/dpaa2/dpaa2_rxtx.c | 119 ---
3 files changed, 79 insertions(+), 47 deletions(-)
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 698fb84..797e71d 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -391,7 +391,7 @@
RTE_PTYPE_UNKNOWN
};
- if (dev->rx_pkt_burst == dpaa2_dev_rx)
+ if (dev->rx_pkt_burst == dpaa2_dev_prefetch_rx)
return ptypes;
return NULL;
}
@@ -886,7 +886,7 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
eth_dev->dev_ops = &dpaa2_ethdev_ops;
eth_dev->data->drv_name = rte_dpaa2_pmd.driver.name;
- eth_dev->rx_pkt_burst = dpaa2_dev_rx;
+ eth_dev->rx_pkt_burst = dpaa2_dev_prefetch_rx;
eth_dev->tx_pkt_burst = dpaa2_dev_tx;
rte_fslmc_vfio_dmamap();
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.h b/drivers/net/dpaa2/dpaa2_ethdev.h
index 7196398..7fa7e7d 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.h
+++ b/drivers/net/dpaa2/dpaa2_ethdev.h
@@ -77,7 +77,8 @@ int dpaa2_remove_flow_dist(struct rte_eth_dev *eth_dev,
int dpaa2_attach_bp_list(struct dpaa2_dev_priv *priv, void *blist);
-uint16_t dpaa2_dev_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts);
+uint16_t dpaa2_dev_prefetch_rx(void *queue, struct rte_mbuf **bufs,
+ uint16_t nb_pkts);
uint16_t dpaa2_dev_tx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts);
#endif /* _DPAA2_ETHDEV_H */
diff --git a/drivers/net/dpaa2/dpaa2_rxtx.c b/drivers/net/dpaa2/dpaa2_rxtx.c
index c5d49cb..f1e866e 100644
--- a/drivers/net/dpaa2/dpaa2_rxtx.c
+++ b/drivers/net/dpaa2/dpaa2_rxtx.c
@@ -242,17 +242,18 @@ static inline int __attribute__((hot))
}
uint16_t
-dpaa2_dev_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
+dpaa2_dev_prefetch_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
{
- /* Function is responsible to receive frames for a given device and VQ*/
+ /* Function receive frames for a given device and VQ*/
struct dpaa2_queue *dpaa2_q = (struct dpaa2_queue *)queue;
struct qbman_result *dq_storage;
uint32_t fqid = dpaa2_q->fqid;
int ret, num_rx = 0;
uint8_t is_last = 0, status;
struct qbman_swp *swp;
- const struct qbman_fd *fd;
+ const struct qbman_fd *fd[DPAA2_DQRR_RING_SIZE];
struct qbman_pull_desc pulldesc;
+ struct queue_storage_info_t *q_storage = dpaa2_q->q_storage;
struct rte_eth_dev *dev = dpaa2_q->dev;
if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
@@ -263,44 +264,51 @@ static inline int __attribute__((hot))
}
}
swp = DPAA2_PER_LCORE_PORTAL;
- dq_storage = dpaa2_q->q_storage->dq_storage[0];
-
- qbman_pull_desc_clear(&pulldesc);
- qbman_pull_desc_set_numframes(&pulldesc,
- (nb_pkts > DPAA2_DQRR_RING_SIZE) ?
- DPAA2_DQRR_RING_SIZE : nb_pkts);
- qbman_pull_desc_set_fq(&pulldesc, fqid);
- /* todo optimization - we can have dq_storage_phys available*/
- qbman_pull_desc_set_storage(&pulldesc, dq_storage,
+ if (!q_storage->active_dqs) {
+ q_storage->toggle = 0;
+ dq_storage = q_storage->dq_storage[q_storage->toggle];
+ qbman_pull_desc_clear(&pulldesc);
+ qbman_pull_desc_set_numframes(&pulldesc,
+ (nb_pkts > DPAA2_DQRR_RING_SIZE) ?
+ DPAA2_DQRR_RING_SIZE : nb_pkts);
+ qbman_pull_desc_set_fq(&pulldesc, fqid);
+ qbman_pull_desc_set_storage(&pulldesc, dq_storage,
(dma_addr_t)(DPAA2_VADDR_TO_IOVA(dq_storage)), 1);
-
- /*Issue a volatile dequeue command. */
- while (1) {
- if (qbman_swp_pull(swp, &pulldesc)) {
- PMD_RX_LOG(ERR, "VDQ command is not issued."
- "QBMAN is busy\n");
- /* Portal was busy, try again */
- continue;
+ if (check_swp_active_dqs(DPAA2_PER_LCORE_DPIO->index)) {
+ while (!qbman_check_command_complete(swp,
+ get_swp_active_dqs(DPAA2_PER_LCORE_DPIO->index)))
[dpdk-dev] [PATCH v2 11/20] net/dpaa2: add support for MAC address filtering
Signed-off-by: Hemant Agrawal
---
doc/guides/nics/features/dpaa2.ini | 1 +
drivers/net/dpaa2/dpaa2_ethdev.c| 81 -
drivers/net/dpaa2/mc/dpni.c | 76 ++
drivers/net/dpaa2/mc/fsl_dpni.h | 45 +
drivers/net/dpaa2/mc/fsl_dpni_cmd.h | 42 +++
5 files changed, 244 insertions(+), 1 deletion(-)
diff --git a/doc/guides/nics/features/dpaa2.ini
b/doc/guides/nics/features/dpaa2.ini
index d43f404..470853c 100644
--- a/doc/guides/nics/features/dpaa2.ini
+++ b/doc/guides/nics/features/dpaa2.ini
@@ -8,6 +8,7 @@ Link status = Y
Queue start/stop = Y
MTU update = Y
Promiscuous mode = Y
+Unicast MAC filter = Y
RSS hash = Y
L3 checksum offload = Y
L4 checksum offload = Y
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 0526e26..5180871 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -672,6 +672,78 @@
return 0;
}
+static int
+dpaa2_dev_add_mac_addr(struct rte_eth_dev *dev,
+ struct ether_addr *addr,
+ __rte_unused uint32_t index,
+ __rte_unused uint32_t pool)
+{
+ int ret;
+ struct dpaa2_dev_priv *priv = dev->data->dev_private;
+ struct fsl_mc_io *dpni = (struct fsl_mc_io *)priv->hw;
+
+ PMD_INIT_FUNC_TRACE();
+
+ if (dpni == NULL) {
+ RTE_LOG(ERR, PMD, "dpni is NULL");
+ return -1;
+ }
+
+ ret = dpni_add_mac_addr(dpni, CMD_PRI_LOW,
+ priv->token, addr->addr_bytes);
+ if (ret)
+ RTE_LOG(ERR, PMD, "error: Adding the MAC ADDR failed:"
+ " err = %d", ret);
+ return 0;
+}
+
+static void
+dpaa2_dev_remove_mac_addr(struct rte_eth_dev *dev,
+ uint32_t index)
+{
+ int ret;
+ struct dpaa2_dev_priv *priv = dev->data->dev_private;
+ struct fsl_mc_io *dpni = (struct fsl_mc_io *)priv->hw;
+ struct rte_eth_dev_data *data = dev->data;
+ struct ether_addr *macaddr;
+
+ PMD_INIT_FUNC_TRACE();
+
+ macaddr = &data->mac_addrs[index];
+
+ if (dpni == NULL) {
+ RTE_LOG(ERR, PMD, "dpni is NULL");
+ return;
+ }
+
+ ret = dpni_remove_mac_addr(dpni, CMD_PRI_LOW,
+ priv->token, macaddr->addr_bytes);
+ if (ret)
+ RTE_LOG(ERR, PMD, "error: Removing the MAC ADDR failed:"
+ " err = %d", ret);
+}
+
+static void
+dpaa2_dev_set_mac_addr(struct rte_eth_dev *dev,
+ struct ether_addr *addr)
+{
+ int ret;
+ struct dpaa2_dev_priv *priv = dev->data->dev_private;
+ struct fsl_mc_io *dpni = (struct fsl_mc_io *)priv->hw;
+
+ PMD_INIT_FUNC_TRACE();
+
+ if (dpni == NULL) {
+ RTE_LOG(ERR, PMD, "dpni is NULL");
+ return;
+ }
+
+ ret = dpni_set_primary_mac_addr(dpni, CMD_PRI_LOW,
+ priv->token, addr->addr_bytes);
+
+ if (ret)
+ RTE_LOG(ERR, PMD, "error: Setting the MAC ADDR failed %d", ret);
+}
static
void dpaa2_dev_stats_get(struct rte_eth_dev *dev,
struct rte_eth_stats *stats)
@@ -720,7 +792,11 @@ void dpaa2_dev_stats_get(struct rte_eth_dev *dev,
if (retcode)
goto err;
- stats->ierrors = value.page_2.ingress_discarded_frames;
+ /* Ingress drop frame count due to configured rules */
+ stats->ierrors = value.page_2.ingress_filtered_frames;
+ /* Ingress drop frame count due to error */
+ stats->ierrors += value.page_2.ingress_discarded_frames;
+
stats->oerrors = value.page_2.egress_discarded_frames;
stats->imissed = value.page_2.ingress_nobuffer_discards;
@@ -822,6 +898,9 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
.rx_queue_release = dpaa2_dev_rx_queue_release,
.tx_queue_setup= dpaa2_dev_tx_queue_setup,
.tx_queue_release = dpaa2_dev_tx_queue_release,
+ .mac_addr_add = dpaa2_dev_add_mac_addr,
+ .mac_addr_remove = dpaa2_dev_remove_mac_addr,
+ .mac_addr_set = dpaa2_dev_set_mac_addr,
};
static int
diff --git a/drivers/net/dpaa2/mc/dpni.c b/drivers/net/dpaa2/mc/dpni.c
index ad33282..38c5f3f 100644
--- a/drivers/net/dpaa2/mc/dpni.c
+++ b/drivers/net/dpaa2/mc/dpni.c
@@ -591,6 +591,82 @@ int dpni_get_primary_mac_addr(struct fsl_mc_io *mc_io,
return 0;
}
+int dpni_add_mac_addr(struct fsl_mc_io *mc_io,
+ uint32_t cmd_flags,
+ uint16_t token,
+ const uint8_t mac_addr[6])
+{
+ struct mc_command cmd = { 0 };
+
+ /* prepare command */
+ cmd.header = mc_encode_cmd_header(DPNI_CMDID_ADD_MAC_ADDR,
+
[dpdk-dev] [PATCH v2 10/20] bus/fslmc: update TAILQ usages in dpaa2 objects
This patch updates the usages of malloc and TAILQ in
dpbp and dpio objects.
Signed-off-by: Hemant Agrawal
---
drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c | 28 +---
drivers/bus/fslmc/portal/dpaa2_hw_dpio.c | 32
2 files changed, 21 insertions(+), 39 deletions(-)
diff --git a/drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c
b/drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c
index 2fb285c..644f01c 100644
--- a/drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c
+++ b/drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c
@@ -53,8 +53,9 @@
#include "portal/dpaa2_hw_pvt.h"
#include "portal/dpaa2_hw_dpio.h"
-TAILQ_HEAD(dpbp_device_list, dpaa2_dpbp_dev);
-static struct dpbp_device_list *dpbp_dev_list; /*!< DPBP device list */
+TAILQ_HEAD(dpbp_dev_list, dpaa2_dpbp_dev);
+static struct dpbp_dev_list dpbp_dev_list
+ = TAILQ_HEAD_INITIALIZER(dpbp_dev_list); /*!< DPBP device list */
int
dpaa2_create_dpbp_device(
@@ -63,19 +64,8 @@
struct dpaa2_dpbp_dev *dpbp_node;
int ret;
- if (!dpbp_dev_list) {
- dpbp_dev_list = malloc(sizeof(struct dpbp_device_list));
- if (!dpbp_dev_list) {
- PMD_INIT_LOG(ERR, "Memory alloc failed in DPBP list\n");
- return -1;
- }
- /* Initialize the DPBP List */
- TAILQ_INIT(dpbp_dev_list);
- }
-
/* Allocate DPAA2 dpbp handle */
- dpbp_node = (struct dpaa2_dpbp_dev *)
- malloc(sizeof(struct dpaa2_dpbp_dev));
+ dpbp_node = rte_malloc(NULL, sizeof(struct dpaa2_dpbp_dev), 0);
if (!dpbp_node) {
PMD_INIT_LOG(ERR, "Memory allocation failed for DPBP Device");
return -1;
@@ -88,7 +78,7 @@
if (ret) {
PMD_INIT_LOG(ERR, "Resource alloc failure with err code: %d",
ret);
- free(dpbp_node);
+ rte_free(dpbp_node);
return -1;
}
@@ -98,14 +88,14 @@
PMD_INIT_LOG(ERR, "Failure cleaning dpbp device with"
" error code %d\n", ret);
dpbp_close(&dpbp_node->dpbp, CMD_PRI_LOW, dpbp_node->token);
- free(dpbp_node);
+ rte_free(dpbp_node);
return -1;
}
dpbp_node->dpbp_id = dpbp_id;
rte_atomic16_init(&dpbp_node->in_use);
- TAILQ_INSERT_HEAD(dpbp_dev_list, dpbp_node, next);
+ TAILQ_INSERT_TAIL(&dpbp_dev_list, dpbp_node, next);
PMD_INIT_LOG(DEBUG, "Buffer pool resource initialized %d", dpbp_id);
@@ -117,7 +107,7 @@ struct dpaa2_dpbp_dev *dpaa2_alloc_dpbp_dev(void)
struct dpaa2_dpbp_dev *dpbp_dev = NULL;
/* Get DPBP dev handle from list using index */
- TAILQ_FOREACH(dpbp_dev, dpbp_dev_list, next) {
+ TAILQ_FOREACH(dpbp_dev, &dpbp_dev_list, next) {
if (dpbp_dev && rte_atomic16_test_and_set(&dpbp_dev->in_use))
break;
}
@@ -130,7 +120,7 @@ void dpaa2_free_dpbp_dev(struct dpaa2_dpbp_dev *dpbp)
struct dpaa2_dpbp_dev *dpbp_dev = NULL;
/* Match DPBP handle and mark it free */
- TAILQ_FOREACH(dpbp_dev, dpbp_dev_list, next) {
+ TAILQ_FOREACH(dpbp_dev, &dpbp_dev_list, next) {
if (dpbp_dev == dpbp) {
rte_atomic16_dec(&dpbp_dev->in_use);
return;
diff --git a/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
b/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
index 203a7d5..73b1808 100644
--- a/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
+++ b/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
@@ -69,8 +69,9 @@
struct swp_active_dqs global_active_dqs_list[NUM_MAX_SWP];
-TAILQ_HEAD(dpio_device_list, dpaa2_dpio_dev);
-static struct dpio_device_list *dpio_dev_list; /*!< DPIO device list */
+TAILQ_HEAD(dpio_dev_list, dpaa2_dpio_dev);
+static struct dpio_dev_list dpio_dev_list
+ = TAILQ_HEAD_INITIALIZER(dpio_dev_list); /*!< DPIO device list */
static uint32_t io_space_count;
/*Stashing Macros default for LS208x*/
@@ -214,7 +215,7 @@ static inline struct dpaa2_dpio_dev
*dpaa2_get_qbman_swp(void)
int ret;
/* Get DPIO dev handle from list using index */
- TAILQ_FOREACH(dpio_dev, dpio_dev_list, next) {
+ TAILQ_FOREACH(dpio_dev, &dpio_dev_list, next) {
if (dpio_dev && rte_atomic16_test_and_set(&dpio_dev->ref_count))
break;
}
@@ -336,18 +337,8 @@ static inline struct dpaa2_dpio_dev
*dpaa2_get_qbman_swp(void)
return -1;
}
- if (!dpio_dev_list) {
- dpio_dev_list = malloc(sizeof(struct dpio_device_list));
- if (!dpio_dev_list) {
- PMD_INIT_LOG(ERR, "Memory alloc failed in DPIO list\n");
- return -1;
- }
-
- /* Initialize the DPIO List */
- TAILQ_INIT(dpi
[dpdk-dev] [PATCH v2 12/20] net/dpaa2: add support for multicast promiscuous mode
Signed-off-by: Hemant Agrawal
---
doc/guides/nics/features/dpaa2.ini | 1 +
drivers/net/dpaa2/dpaa2_ethdev.c| 61 +++--
drivers/net/dpaa2/mc/dpni.c | 41 +
drivers/net/dpaa2/mc/fsl_dpni.h | 27
drivers/net/dpaa2/mc/fsl_dpni_cmd.h | 10 ++
5 files changed, 138 insertions(+), 2 deletions(-)
diff --git a/doc/guides/nics/features/dpaa2.ini
b/doc/guides/nics/features/dpaa2.ini
index 470853c..dddf0e0 100644
--- a/doc/guides/nics/features/dpaa2.ini
+++ b/doc/guides/nics/features/dpaa2.ini
@@ -8,6 +8,7 @@ Link status = Y
Queue start/stop = Y
MTU update = Y
Promiscuous mode = Y
+Allmulticast mode= Y
Unicast MAC filter = Y
RSS hash = Y
L3 checksum offload = Y
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 5180871..22e0474 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -617,7 +617,11 @@
ret = dpni_set_unicast_promisc(dpni, CMD_PRI_LOW, priv->token, true);
if (ret < 0)
- RTE_LOG(ERR, PMD, "Unable to enable promiscuous mode %d", ret);
+ RTE_LOG(ERR, PMD, "Unable to enable U promisc mode %d", ret);
+
+ ret = dpni_set_multicast_promisc(dpni, CMD_PRI_LOW, priv->token, true);
+ if (ret < 0)
+ RTE_LOG(ERR, PMD, "Unable to enable M promisc mode %d", ret);
}
static void
@@ -637,7 +641,58 @@
ret = dpni_set_unicast_promisc(dpni, CMD_PRI_LOW, priv->token, false);
if (ret < 0)
- RTE_LOG(ERR, PMD, "Unable to disable promiscuous mode %d", ret);
+ RTE_LOG(ERR, PMD, "Unable to disable U promisc mode %d", ret);
+
+ if (dev->data->all_multicast == 0) {
+ ret = dpni_set_multicast_promisc(dpni, CMD_PRI_LOW,
+priv->token, false);
+ if (ret < 0)
+ RTE_LOG(ERR, PMD, "Unable to disable M promisc mode %d",
+ ret);
+ }
+}
+
+static void
+dpaa2_dev_allmulticast_enable(
+ struct rte_eth_dev *dev)
+{
+ int ret;
+ struct dpaa2_dev_priv *priv = dev->data->dev_private;
+ struct fsl_mc_io *dpni = (struct fsl_mc_io *)priv->hw;
+
+ PMD_INIT_FUNC_TRACE();
+
+ if (dpni == NULL) {
+ RTE_LOG(ERR, PMD, "dpni is NULL");
+ return;
+ }
+
+ ret = dpni_set_multicast_promisc(dpni, CMD_PRI_LOW, priv->token, true);
+ if (ret < 0)
+ RTE_LOG(ERR, PMD, "Unable to enable multicast mode %d", ret);
+}
+
+static void
+dpaa2_dev_allmulticast_disable(struct rte_eth_dev *dev)
+{
+ int ret;
+ struct dpaa2_dev_priv *priv = dev->data->dev_private;
+ struct fsl_mc_io *dpni = (struct fsl_mc_io *)priv->hw;
+
+ PMD_INIT_FUNC_TRACE();
+
+ if (dpni == NULL) {
+ RTE_LOG(ERR, PMD, "dpni is NULL");
+ return;
+ }
+
+ /* must remain on for all promiscuous */
+ if (dev->data->promiscuous == 1)
+ return;
+
+ ret = dpni_set_multicast_promisc(dpni, CMD_PRI_LOW, priv->token, false);
+ if (ret < 0)
+ RTE_LOG(ERR, PMD, "Unable to disable multicast mode %d", ret);
}
static int
@@ -888,6 +943,8 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
.dev_close= dpaa2_dev_close,
.promiscuous_enable = dpaa2_dev_promiscuous_enable,
.promiscuous_disable = dpaa2_dev_promiscuous_disable,
+ .allmulticast_enable = dpaa2_dev_allmulticast_enable,
+ .allmulticast_disable = dpaa2_dev_allmulticast_disable,
.link_update = dpaa2_dev_link_update,
.stats_get = dpaa2_dev_stats_get,
.stats_reset = dpaa2_dev_stats_reset,
diff --git a/drivers/net/dpaa2/mc/dpni.c b/drivers/net/dpaa2/mc/dpni.c
index 38c5f3f..54a1568 100644
--- a/drivers/net/dpaa2/mc/dpni.c
+++ b/drivers/net/dpaa2/mc/dpni.c
@@ -509,6 +509,47 @@ int dpni_get_max_frame_length(struct fsl_mc_io *mc_io,
return 0;
}
+int dpni_set_multicast_promisc(struct fsl_mc_io *mc_io,
+ uint32_t cmd_flags,
+ uint16_t token,
+ int en)
+{
+ struct mc_command cmd = { 0 };
+
+ /* prepare command */
+ cmd.header = mc_encode_cmd_header(DPNI_CMDID_SET_MCAST_PROMISC,
+ cmd_flags,
+ token);
+ DPNI_CMD_SET_MULTICAST_PROMISC(cmd, en);
+
+ /* send command to mc*/
+ return mc_send_command(mc_io, &cmd);
+}
+
+int dpni_get_multicast_promisc(struct fsl_mc_io *mc_io,
+ uint32_t cmd_flags,
+ uint16_t token,
+ int *en)
+{
+ struct mc_command cmd = { 0 };
+ int err;
+
+ /* prepa
[dpdk-dev] [PATCH v2 13/20] net/dpaa2: add support for VLAN filter and offload
Signed-off-by: Hemant Agrawal
---
doc/guides/nics/features/dpaa2.ini | 2 ++
drivers/net/dpaa2/dpaa2_ethdev.c| 55 +++
drivers/net/dpaa2/mc/dpni.c | 66 +
drivers/net/dpaa2/mc/fsl_dpni.h | 54 ++
drivers/net/dpaa2/mc/fsl_dpni_cmd.h | 17 ++
5 files changed, 194 insertions(+)
diff --git a/doc/guides/nics/features/dpaa2.ini
b/doc/guides/nics/features/dpaa2.ini
index dddf0e0..4b1e000 100644
--- a/doc/guides/nics/features/dpaa2.ini
+++ b/doc/guides/nics/features/dpaa2.ini
@@ -11,6 +11,8 @@ Promiscuous mode = Y
Allmulticast mode= Y
Unicast MAC filter = Y
RSS hash = Y
+VLAN filter = Y
+VLAN offload = Y
L3 checksum offload = Y
L4 checksum offload = Y
Packet type parsing = Y
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 22e0474..46532b2 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -108,6 +108,56 @@
return 0;
}
+static int
+dpaa2_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
+{
+ int ret;
+ struct dpaa2_dev_priv *priv = dev->data->dev_private;
+ struct fsl_mc_io *dpni = priv->hw;
+
+ PMD_INIT_FUNC_TRACE();
+
+ if (dpni == NULL) {
+ RTE_LOG(ERR, PMD, "dpni is NULL");
+ return -1;
+ }
+
+ if (on)
+ ret = dpni_add_vlan_id(dpni, CMD_PRI_LOW,
+ priv->token, vlan_id);
+ else
+ ret = dpni_remove_vlan_id(dpni, CMD_PRI_LOW,
+ priv->token, vlan_id);
+
+ if (ret < 0)
+ PMD_DRV_LOG(ERR, "ret = %d Unable to add/rem vlan %d hwid =%d",
+ ret, vlan_id, priv->hw_id);
+
+ return ret;
+}
+
+static void
+dpaa2_vlan_offload_set(struct rte_eth_dev *dev, int mask)
+{
+ struct dpaa2_dev_priv *priv = dev->data->dev_private;
+ struct fsl_mc_io *dpni = priv->hw;
+ int ret;
+
+ PMD_INIT_FUNC_TRACE();
+
+ if (mask & ETH_VLAN_FILTER_MASK) {
+ if (dev->data->dev_conf.rxmode.hw_vlan_filter)
+ ret = dpni_enable_vlan_filter(dpni, CMD_PRI_LOW,
+ priv->token, true);
+ else
+ ret = dpni_enable_vlan_filter(dpni, CMD_PRI_LOW,
+ priv->token, false);
+ if (ret < 0)
+ RTE_LOG(ERR, PMD, "Unable to set vlan filter ret = %d",
+ ret);
+ }
+}
+
static void
dpaa2_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
{
@@ -542,6 +592,9 @@
"code = %d\n", ret);
return ret;
}
+ /* VLAN Offload Settings */
+ if (priv->max_vlan_filters)
+ dpaa2_vlan_offload_set(dev, ETH_VLAN_FILTER_MASK);
return 0;
}
@@ -951,6 +1004,8 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
.dev_infos_get = dpaa2_dev_info_get,
.dev_supported_ptypes_get = dpaa2_supported_ptypes_get,
.mtu_set = dpaa2_dev_mtu_set,
+ .vlan_filter_set = dpaa2_vlan_filter_set,
+ .vlan_offload_set = dpaa2_vlan_offload_set,
.rx_queue_setup= dpaa2_dev_rx_queue_setup,
.rx_queue_release = dpaa2_dev_rx_queue_release,
.tx_queue_setup= dpaa2_dev_tx_queue_setup,
diff --git a/drivers/net/dpaa2/mc/dpni.c b/drivers/net/dpaa2/mc/dpni.c
index 54a1568..e92fe06 100644
--- a/drivers/net/dpaa2/mc/dpni.c
+++ b/drivers/net/dpaa2/mc/dpni.c
@@ -708,6 +708,72 @@ int dpni_get_port_mac_addr(struct fsl_mc_io *mc_io,
return 0;
}
+int dpni_enable_vlan_filter(struct fsl_mc_io *mc_io,
+ uint32_t cmd_flags,
+ uint16_t token,
+ int en)
+{
+ struct mc_command cmd = { 0 };
+
+ /* prepare command */
+ cmd.header = mc_encode_cmd_header(DPNI_CMDID_ENABLE_VLAN_FILTER,
+ cmd_flags,
+ token);
+ DPNI_CMD_ENABLE_VLAN_FILTER(cmd, en);
+
+ /* send command to mc*/
+ return mc_send_command(mc_io, &cmd);
+}
+
+int dpni_add_vlan_id(struct fsl_mc_io *mc_io,
+uint32_t cmd_flags,
+uint16_t token,
+uint16_t vlan_id)
+{
+ struct mc_command cmd = { 0 };
+
+ /* prepare command */
+ cmd.header = mc_encode_cmd_header(DPNI_CMDID_ADD_VLAN_ID,
+ cmd_flags,
+ token);
+ DPNI_CMD_ADD_VLAN_ID(cmd, vlan_id);
+
+ /* send command to mc*/
+ return mc_send_command(mc_io, &cmd);
+}
+
+int dpni_remove_vlan_id(struct fsl_mc_io *
[dpdk-dev] [PATCH v2 14/20] net/dpaa2: add support for VLAN strip
Signed-off-by: Hemant Agrawal --- drivers/net/dpaa2/dpaa2_rxtx.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/net/dpaa2/dpaa2_rxtx.c b/drivers/net/dpaa2/dpaa2_rxtx.c index 0670ae3..d7e18ad 100644 --- a/drivers/net/dpaa2/dpaa2_rxtx.c +++ b/drivers/net/dpaa2/dpaa2_rxtx.c @@ -330,6 +330,9 @@ static inline int __attribute__((hot)) if (dev->data->dev_conf.rxmode.hw_vlan_strip) rte_vlan_strip(bufs[num_rx]); + if (dev->data->dev_conf.rxmode.hw_vlan_strip) + rte_vlan_strip(bufs[num_rx]); + dq_storage++; num_rx++; } -- 1.9.1
[dpdk-dev] [PATCH v2 16/20] net/dpaa2: add support for flow control
Signed-off-by: Hemant Agrawal
---
doc/guides/nics/features/dpaa2.ini | 1 +
drivers/net/dpaa2/dpaa2_ethdev.c| 142
drivers/net/dpaa2/mc/dpni.c | 18 +
drivers/net/dpaa2/mc/fsl_dpni.h | 24 ++
drivers/net/dpaa2/mc/fsl_dpni_cmd.h | 8 ++
5 files changed, 193 insertions(+)
diff --git a/doc/guides/nics/features/dpaa2.ini
b/doc/guides/nics/features/dpaa2.ini
index 4b1e000..720c35b 100644
--- a/doc/guides/nics/features/dpaa2.ini
+++ b/doc/guides/nics/features/dpaa2.ini
@@ -12,6 +12,7 @@ Allmulticast mode= Y
Unicast MAC filter = Y
RSS hash = Y
VLAN filter = Y
+Flow control = Y
VLAN offload = Y
L3 checksum offload = Y
L4 checksum offload = Y
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index fa9db17..edff04a 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -1105,6 +1105,146 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
return ret;
}
+static int
+dpaa2_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
+{
+ int ret = -EINVAL;
+ struct dpaa2_dev_priv *priv;
+ struct fsl_mc_io *dpni;
+ struct dpni_link_state state = {0};
+
+ PMD_INIT_FUNC_TRACE();
+
+ priv = dev->data->dev_private;
+ dpni = (struct fsl_mc_io *)priv->hw;
+
+ if (dpni == NULL || fc_conf == NULL) {
+ RTE_LOG(ERR, PMD, "device not configured");
+ return ret;
+ }
+
+ ret = dpni_get_link_state(dpni, CMD_PRI_LOW, priv->token, &state);
+ if (ret) {
+ RTE_LOG(ERR, PMD, "error: dpni_get_link_state %d", ret);
+ return ret;
+ }
+
+ memset(fc_conf, 0, sizeof(struct rte_eth_fc_conf));
+ if (state.options & DPNI_LINK_OPT_PAUSE) {
+ /* DPNI_LINK_OPT_PAUSE set
+* if ASYM_PAUSE not set,
+* RX Side flow control (handle received Pause frame)
+* TX side flow control (send Pause frame)
+* if ASYM_PAUSE set,
+* RX Side flow control (handle received Pause frame)
+* No TX side flow control (send Pause frame disabled)
+*/
+ if (!(state.options & DPNI_LINK_OPT_ASYM_PAUSE))
+ fc_conf->mode = RTE_FC_FULL;
+ else
+ fc_conf->mode = RTE_FC_RX_PAUSE;
+ } else {
+ /* DPNI_LINK_OPT_PAUSE not set
+* if ASYM_PAUSE set,
+* TX side flow control (send Pause frame)
+* No RX side flow control (No action on pause frame rx)
+* if ASYM_PAUSE not set,
+* Flow control disabled
+*/
+ if (state.options & DPNI_LINK_OPT_ASYM_PAUSE)
+ fc_conf->mode = RTE_FC_TX_PAUSE;
+ else
+ fc_conf->mode = RTE_FC_NONE;
+ }
+
+ return ret;
+}
+
+static int
+dpaa2_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
+{
+ int ret = -EINVAL;
+ struct dpaa2_dev_priv *priv;
+ struct fsl_mc_io *dpni;
+ struct dpni_link_state state = {0};
+ struct dpni_link_cfg cfg = {0};
+
+ PMD_INIT_FUNC_TRACE();
+
+ priv = dev->data->dev_private;
+ dpni = (struct fsl_mc_io *)priv->hw;
+
+ if (dpni == NULL) {
+ RTE_LOG(ERR, PMD, "dpni is NULL");
+ return ret;
+ }
+
+ /* It is necessary to obtain the current state before setting fc_conf
+* as MC would return error in case rate, autoneg or duplex values are
+* different.
+*/
+ ret = dpni_get_link_state(dpni, CMD_PRI_LOW, priv->token, &state);
+ if (ret) {
+ RTE_LOG(ERR, PMD, "Unable to get link state (err=%d)", ret);
+ return -1;
+ }
+
+ /* Disable link before setting configuration */
+ dpaa2_dev_set_link_down(dev);
+
+ /* Based on fc_conf, update cfg */
+ cfg.rate = state.rate;
+ cfg.options = state.options;
+
+ /* update cfg with fc_conf */
+ switch (fc_conf->mode) {
+ case RTE_FC_FULL:
+ /* Full flow control;
+* OPT_PAUSE set, ASYM_PAUSE not set
+*/
+ cfg.options |= DPNI_LINK_OPT_PAUSE;
+ cfg.options &= ~DPNI_LINK_OPT_ASYM_PAUSE;
+ case RTE_FC_TX_PAUSE:
+ /* Enable RX flow control
+* OPT_PAUSE not set;
+* ASYM_PAUSE set;
+*/
+ cfg.options |= DPNI_LINK_OPT_ASYM_PAUSE;
+ cfg.options &= ~DPNI_LINK_OPT_PAUSE;
+ break;
+ case RTE_FC_RX_PAUSE:
+ /* Enable TX Flow control
+* OPT_PAUSE set
+* ASYM_PAUSE set
+*/
+
[dpdk-dev] [PATCH v2 17/20] net/dpaa2: configure jumbo frames
Signed-off-by: Hemant Agrawal
---
doc/guides/nics/features/dpaa2.ini | 1 +
drivers/net/dpaa2/dpaa2_ethdev.c | 19 +++
2 files changed, 20 insertions(+)
diff --git a/doc/guides/nics/features/dpaa2.ini
b/doc/guides/nics/features/dpaa2.ini
index 720c35b..48422f0 100644
--- a/doc/guides/nics/features/dpaa2.ini
+++ b/doc/guides/nics/features/dpaa2.ini
@@ -6,6 +6,7 @@
[Features]
Link status = Y
Queue start/stop = Y
+Jumbo frame = Y
MTU update = Y
Promiscuous mode = Y
Allmulticast mode= Y
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index edff04a..31a2358 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -57,6 +57,7 @@
static int dpaa2_dev_uninit(struct rte_eth_dev *eth_dev);
static int dpaa2_dev_set_link_up(struct rte_eth_dev *dev);
static int dpaa2_dev_set_link_down(struct rte_eth_dev *dev);
+static int dpaa2_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
/**
* Atomically reads the link status information from global
@@ -278,6 +279,19 @@
PMD_INIT_FUNC_TRACE();
+ if (eth_conf->rxmode.jumbo_frame == 1) {
+ if (eth_conf->rxmode.max_rx_pkt_len <= DPAA2_MAX_RX_PKT_LEN) {
+ ret = dpaa2_dev_mtu_set(dev,
+ eth_conf->rxmode.max_rx_pkt_len);
+ if (ret) {
+ PMD_INIT_LOG(ERR, "unable to set mtu."
+"please check queue config\n");
+ return ret;
+ }
+ } else
+ return -1;
+ }
+
/* Check for correct configuration */
if (eth_conf->rxmode.mq_mode != ETH_MQ_RX_RSS &&
data->nb_rx_queues > 1) {
@@ -778,6 +792,11 @@
if ((mtu < ETHER_MIN_MTU) || (frame_size > DPAA2_MAX_RX_PKT_LEN))
return -EINVAL;
+ if (frame_size > ETHER_MAX_LEN)
+ dev->data->dev_conf.rxmode.jumbo_frame = 1;
+ else
+ dev->data->dev_conf.rxmode.jumbo_frame = 0;
+
/* Set the Max Rx frame length as 'mtu' +
* Maximum Ethernet header length
*/
--
1.9.1
[dpdk-dev] [PATCH v2 15/20] net/dpaa2: add link status config support
Signed-off-by: Hemant Agrawal
---
drivers/net/dpaa2/dpaa2_ethdev.c | 118 +++
drivers/net/dpaa2/dpaa2_ethdev.h | 2 +-
drivers/net/dpaa2/dpaa2_rxtx.c | 25 +
3 files changed, 144 insertions(+), 1 deletion(-)
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 46532b2..fa9db17 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -55,6 +55,8 @@
static struct rte_dpaa2_driver rte_dpaa2_pmd;
static int dpaa2_dev_uninit(struct rte_eth_dev *eth_dev);
+static int dpaa2_dev_set_link_up(struct rte_eth_dev *dev);
+static int dpaa2_dev_set_link_down(struct rte_eth_dev *dev);
/**
* Atomically reads the link status information from global
@@ -530,6 +532,9 @@
return ret;
}
+ /* Power up the phy. Needed to make the link go Up */
+ dpaa2_dev_set_link_up(dev);
+
ret = dpni_get_qdid(dpni, CMD_PRI_LOW, priv->token,
DPNI_QUEUE_TX, &qdid);
if (ret) {
@@ -613,6 +618,8 @@
PMD_INIT_FUNC_TRACE();
+ dpaa2_dev_set_link_down(dev);
+
ret = dpni_disable(dpni, CMD_PRI_LOW, priv->token);
if (ret) {
PMD_INIT_LOG(ERR, "Failure (ret %d) in disabling dpni %d dev\n",
@@ -632,6 +639,7 @@
struct dpaa2_dev_priv *priv = dev->data->dev_private;
struct fsl_mc_io *dpni = (struct fsl_mc_io *)priv->hw;
int i, ret;
+ struct rte_eth_link link;
struct dpaa2_queue *dpaa2_q;
PMD_INIT_FUNC_TRACE();
@@ -651,6 +659,9 @@
" error code %d\n", ret);
return;
}
+
+ memset(&link, 0, sizeof(link));
+ dpaa2_dev_atomic_write_link_status(dev, &link);
}
static void
@@ -989,6 +1000,111 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
return 0;
}
+/**
+ * Toggle the DPNI to enable, if not already enabled.
+ * This is not strictly PHY up/down - it is more of logical toggling.
+ */
+static int
+dpaa2_dev_set_link_up(struct rte_eth_dev *dev)
+{
+ int ret = -EINVAL;
+ struct dpaa2_dev_priv *priv;
+ struct fsl_mc_io *dpni;
+ int en = 0;
+
+ PMD_INIT_FUNC_TRACE();
+
+ priv = dev->data->dev_private;
+ dpni = (struct fsl_mc_io *)priv->hw;
+
+ if (dpni == NULL) {
+ RTE_LOG(ERR, PMD, "Device has not yet been configured");
+ return ret;
+ }
+
+ /* Check if DPNI is currently enabled */
+ ret = dpni_is_enabled(dpni, CMD_PRI_LOW, priv->token, &en);
+ if (ret) {
+ /* Unable to obtain dpni status; Not continuing */
+ PMD_DRV_LOG(ERR, "Interface Link UP failed (%d)", ret);
+ return -EINVAL;
+ }
+
+ /* Enable link if not already enabled */
+ if (!en) {
+ ret = dpni_enable(dpni, CMD_PRI_LOW, priv->token);
+ if (ret) {
+ PMD_DRV_LOG(ERR, "Interface Link UP failed (%d)", ret);
+ return -EINVAL;
+ }
+ }
+ /* changing tx burst function to start enqueues */
+ dev->tx_pkt_burst = dpaa2_dev_tx;
+ dev->data->dev_link.link_status = 1;
+
+ PMD_DRV_LOG(INFO, "Port %d Link UP successful", dev->data->port_id);
+ return ret;
+}
+
+/**
+ * Toggle the DPNI to disable, if not already disabled.
+ * This is not strictly PHY up/down - it is more of logical toggling.
+ */
+static int
+dpaa2_dev_set_link_down(struct rte_eth_dev *dev)
+{
+ int ret = -EINVAL;
+ struct dpaa2_dev_priv *priv;
+ struct fsl_mc_io *dpni;
+ int dpni_enabled = 0;
+ int retries = 10;
+
+ PMD_INIT_FUNC_TRACE();
+
+ priv = dev->data->dev_private;
+ dpni = (struct fsl_mc_io *)priv->hw;
+
+ if (dpni == NULL) {
+ RTE_LOG(ERR, PMD, "Device has not yet been configured");
+ return ret;
+ }
+
+ /*changing tx burst function to avoid any more enqueues */
+ dev->tx_pkt_burst = dummy_dev_tx;
+
+ /* Loop while dpni_disable() attempts to drain the egress FQs
+* and confirm them back to us.
+*/
+ do {
+ ret = dpni_disable(dpni, 0, priv->token);
+ if (ret) {
+ PMD_DRV_LOG(ERR, "dpni disable failed (%d)", ret);
+ return ret;
+ }
+ ret = dpni_is_enabled(dpni, 0, priv->token, &dpni_enabled);
+ if (ret) {
+ PMD_DRV_LOG(ERR, "dpni_is_enabled failed (%d)", ret);
+ return ret;
+ }
+ if (dpni_enabled)
+ /* Allow the MC some slack */
+ rte_delay_us(100 * 1000);
+ } while (dpni_enabled && --retries);
+
+ if (!retries) {
+ PMD_DRV_LOG(WARNING, "Retry count exceeded disabling DPNI\n");
+ /* todo- we may have to
[dpdk-dev] [PATCH v2 18/20] bus/fslmc: add support to detect soc version
This is required to configure the stashing destination
based on the cluster size.
Signed-off-by: Hemant Agrawal
---
drivers/bus/fslmc/Makefile | 1 +
drivers/bus/fslmc/mc/dpmng.c| 81 +++
drivers/bus/fslmc/mc/fsl_dpmng.h| 99 +
drivers/bus/fslmc/mc/fsl_dpmng_cmd.h| 54
drivers/bus/fslmc/portal/dpaa2_hw_dpio.c| 23 ++-
drivers/bus/fslmc/rte_bus_fslmc_version.map | 2 +
6 files changed, 258 insertions(+), 2 deletions(-)
create mode 100644 drivers/bus/fslmc/mc/dpmng.c
create mode 100644 drivers/bus/fslmc/mc/fsl_dpmng.h
create mode 100644 drivers/bus/fslmc/mc/fsl_dpmng_cmd.h
diff --git a/drivers/bus/fslmc/Makefile b/drivers/bus/fslmc/Makefile
index 973d279..7d9f4b6 100644
--- a/drivers/bus/fslmc/Makefile
+++ b/drivers/bus/fslmc/Makefile
@@ -63,6 +63,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_FSLMC_BUS) += \
qbman/qbman_portal.c
SRCS-$(CONFIG_RTE_LIBRTE_FSLMC_BUS) += \
+ mc/dpmng.c \
mc/dpbp.c \
mc/dpio.c \
mc/mc_sys.c
diff --git a/drivers/bus/fslmc/mc/dpmng.c b/drivers/bus/fslmc/mc/dpmng.c
new file mode 100644
index 000..7fe1787
--- /dev/null
+++ b/drivers/bus/fslmc/mc/dpmng.c
@@ -0,0 +1,81 @@
+/* Copyright 2013-2015 Freescale Semiconductor Inc.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ * * Neither the name of the above-listed copyright holders nor the
+ * names of any contributors may be used to endorse or promote products
+ * derived from this software without specific prior written permission.
+ *
+ *
+ * ALTERNATIVELY, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") as published by the Free Software
+ * Foundation, either version 2 of that License or (at your option) any
+ * later version.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+#include
+#include
+#include
+#include
+
+int mc_get_version(struct fsl_mc_io *mc_io,
+ uint32_t cmd_flags,
+ struct mc_version *mc_ver_info)
+{
+ struct mc_command cmd = { 0 };
+ int err;
+
+ /* prepare command */
+ cmd.header = mc_encode_cmd_header(DPMNG_CMDID_GET_VERSION,
+ cmd_flags,
+ 0);
+
+ /* send command to mc*/
+ err = mc_send_command(mc_io, &cmd);
+ if (err)
+ return err;
+
+ /* retrieve response parameters */
+ DPMNG_RSP_GET_VERSION(cmd, mc_ver_info);
+
+ return 0;
+}
+
+int mc_get_soc_version(struct fsl_mc_io *mc_io,
+ uint32_t cmd_flags,
+ struct mc_soc_version *mc_platform_info)
+{
+ struct mc_command cmd = { 0 };
+ int err;
+
+ /* prepare command */
+ cmd.header = mc_encode_cmd_header(DPMNG_CMDID_GET_SOC_VERSION,
+ cmd_flags,
+ 0);
+
+ /* send command to mc*/
+ err = mc_send_command(mc_io, &cmd);
+ if (err)
+ return err;
+
+ /* retrieve response parameters */
+ DPMNG_RSP_GET_SOC_VERSION(cmd, mc_platform_info);
+
+ return 0;
+}
diff --git a/drivers/bus/fslmc/mc/fsl_dpmng.h b/drivers/bus/fslmc/mc/fsl_dpmng.h
new file mode 100644
index 000..7dbc825
--- /dev/null
+++ b/drivers/bus/fslmc/mc/fsl_dpmng.h
@@ -0,0 +1,99 @@
+/* Copyright 2013-2015 Freescale Semiconductor Inc.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *
[dpdk-dev] [PATCH v2 20/20] bus/fslmc: reducing the debug log messages
Signed-off-by: Hemant Agrawal
---
drivers/bus/fslmc/fslmc_vfio.c | 13 +++--
drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c | 2 +-
drivers/bus/fslmc/portal/dpaa2_hw_dpio.c | 11 +--
3 files changed, 5 insertions(+), 21 deletions(-)
diff --git a/drivers/bus/fslmc/fslmc_vfio.c b/drivers/bus/fslmc/fslmc_vfio.c
index 5d4ac67..4ecd568 100644
--- a/drivers/bus/fslmc/fslmc_vfio.c
+++ b/drivers/bus/fslmc/fslmc_vfio.c
@@ -132,7 +132,6 @@ static int vfio_connect_container(struct fslmc_vfio_group
*vfio_group)
for (i = 0; i < VFIO_MAX_CONTAINERS; i++) {
if (vfio_containers[i].used)
continue;
- FSLMC_VFIO_LOG(DEBUG, "Unused container at index %d", i);
container = &vfio_containers[i];
}
if (!container) {
@@ -245,7 +244,7 @@ int rte_fslmc_vfio_dmamap(void)
FSLMC_VFIO_LOG(DEBUG, "-->Initial SHM Virtual ADDR %llX",
dma_map.vaddr);
- FSLMC_VFIO_LOG(DEBUG, "-> DMA size 0x%llX\n", dma_map.size);
+ FSLMC_VFIO_LOG(DEBUG, "-> DMA size 0x%llX", dma_map.size);
ret = ioctl(group->container->fd, VFIO_IOMMU_MAP_DMA,
&dma_map);
if (ret) {
@@ -253,8 +252,6 @@ int rte_fslmc_vfio_dmamap(void)
"(errno = %d)", errno);
return ret;
}
- FSLMC_VFIO_LOG(DEBUG, "-> dma_map.vaddr = 0x%llX",
-dma_map.vaddr);
}
/* TODO - This is a W.A. as VFIO currently does not add the mapping of
@@ -441,8 +438,6 @@ int fslmc_vfio_process_group(void)
goto FAILURE;
}
- FSLMC_VFIO_LOG(DEBUG, "DPAA2 MC has VIR_ADD = %ld", v_addr);
-
rte_mcp_ptr_list[0] = (void *)v_addr;
d = opendir(path);
@@ -452,7 +447,6 @@ int fslmc_vfio_process_group(void)
}
i = 0;
- FSLMC_VFIO_LOG(DEBUG, "DPAA2 - Parsing devices:");
/* Parsing each object and initiating them*/
while ((dir = readdir(d)) != NULL) {
if (dir->d_type != DT_LNK)
@@ -469,7 +463,6 @@ int fslmc_vfio_process_group(void)
object_type = strtok(dir->d_name, ".");
temp_obj = strtok(NULL, ".");
sscanf(temp_obj, "%d", &object_id);
- FSLMC_VFIO_LOG(DEBUG, " - %s ", dev_name);
/* getting the device fd*/
dev_fd = ioctl(group->fd, VFIO_GROUP_GET_DEVICE_FD, dev_name);
@@ -505,7 +498,7 @@ int fslmc_vfio_process_group(void)
dev->dev_type = (strcmp(object_type, "dpseci")) ?
DPAA2_MC_DPNI_DEVID : DPAA2_MC_DPSECI_DEVID;
- FSLMC_VFIO_LOG(DEBUG, "DPAA2: Added [%s-%d]\n",
+ FSLMC_VFIO_LOG(DEBUG, "DPAA2: Added [%s-%d]",
object_type, object_id);
fslmc_bus_add_device(dev);
@@ -529,7 +522,7 @@ int fslmc_vfio_process_group(void)
if (ret)
FSLMC_VFIO_LOG(DEBUG, "Error in affining qbman swp %d", ret);
- FSLMC_VFIO_LOG(DEBUG, "DPAA2: Added dpbp_count = %d dpio_count=%d\n",
+ FSLMC_VFIO_LOG(DEBUG, "DPAA2: Added dpbp_count = %d dpio_count=%d",
dpbp_count, dpio_count);
return 0;
diff --git a/drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c
b/drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c
index 644f01c..a665ec5 100644
--- a/drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c
+++ b/drivers/bus/fslmc/portal/dpaa2_hw_dpbp.c
@@ -97,7 +97,7 @@
TAILQ_INSERT_TAIL(&dpbp_dev_list, dpbp_node, next);
- PMD_INIT_LOG(DEBUG, "Buffer pool resource initialized %d", dpbp_id);
+ PMD_INIT_LOG(DEBUG, "DPAA2: Added [dpbp-%d]", dpbp_id);
return 0;
}
diff --git a/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
b/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
index 530da5d..b56b961 100644
--- a/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
+++ b/drivers/bus/fslmc/portal/dpaa2_hw_dpio.c
@@ -155,8 +155,6 @@
}
PMD_INIT_LOG(DEBUG, "Qbman Portal ID %d", attr.qbman_portal_id);
- PMD_INIT_LOG(DEBUG, "Portal CE adr 0x%lX", attr.qbman_portal_ce_offset);
- PMD_INIT_LOG(DEBUG, "Portal CI adr 0x%lX", attr.qbman_portal_ci_offset);
/* Configure & setup SW portal */
p_des.block = NULL;
@@ -174,8 +172,6 @@
return -1;
}
- PMD_INIT_LOG(DEBUG, "QBMan SW Portal 0x%p\n", dpio_dev->sw_portal);
-
return 0;
}
@@ -363,7 +359,6 @@ static inline struct dpaa2_dpio_dev
*dpaa2_get_qbman_swp(void)
return -1;
}
- PMD_DRV_LOG(INFO, "\t Aloocated DPIO [%p]", dpio_dev);
dpio_dev->dpio = NULL;
dpio_dev->hw_id = object_id;
dpio_dev->vfio_fd = vdev->fd;
@@ -378,8 +373,6 @@ static inline struct dpaa2_dpio_dev
*dpaa2_get_qbman_swp(void)
[dpdk-dev] [PATCH v2 19/20] net/dpaa2: add support for Firmware Version get
Signed-off-by: Hemant Agrawal
---
drivers/net/dpaa2/dpaa2_ethdev.c | 36 +++-
1 file changed, 35 insertions(+), 1 deletion(-)
diff --git a/drivers/net/dpaa2/dpaa2_ethdev.c b/drivers/net/dpaa2/dpaa2_ethdev.c
index 31a2358..d64ec48 100644
--- a/drivers/net/dpaa2/dpaa2_ethdev.c
+++ b/drivers/net/dpaa2/dpaa2_ethdev.c
@@ -50,7 +50,7 @@
#include
#include
#include
-
+#include
#include "dpaa2_ethdev.h"
static struct rte_dpaa2_driver rte_dpaa2_pmd;
@@ -161,6 +161,39 @@
}
}
+static int
+dpaa2_fw_version_get(struct rte_eth_dev *dev,
+char *fw_version,
+size_t fw_size)
+{
+ int ret;
+ struct dpaa2_dev_priv *priv = dev->data->dev_private;
+ struct fsl_mc_io *dpni = priv->hw;
+ struct mc_soc_version mc_plat_info = {0};
+ struct mc_version mc_ver_info = {0};
+
+ PMD_INIT_FUNC_TRACE();
+
+ if (mc_get_soc_version(dpni, CMD_PRI_LOW, &mc_plat_info))
+ RTE_LOG(WARNING, PMD, "\tmc_get_soc_version failed\n");
+
+ if (mc_get_version(dpni, CMD_PRI_LOW, &mc_ver_info))
+ RTE_LOG(WARNING, PMD, "\tmc_get_version failed\n");
+
+ ret = snprintf(fw_version, fw_size,
+ "%x-%d.%d.%d",
+ mc_plat_info.svr,
+ mc_ver_info.major,
+ mc_ver_info.minor,
+ mc_ver_info.revision);
+
+ ret += 1; /* add the size of '\0' */
+ if (fw_size < (uint32_t)ret)
+ return ret;
+ else
+ return 0;
+}
+
static void
dpaa2_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
{
@@ -1278,6 +1311,7 @@ void dpaa2_dev_stats_reset(struct rte_eth_dev *dev)
.link_update = dpaa2_dev_link_update,
.stats_get = dpaa2_dev_stats_get,
.stats_reset = dpaa2_dev_stats_reset,
+ .fw_version_get= dpaa2_fw_version_get,
.dev_infos_get = dpaa2_dev_info_get,
.dev_supported_ptypes_get = dpaa2_supported_ptypes_get,
.mtu_set = dpaa2_dev_mtu_set,
--
1.9.1
Re: [dpdk-dev] [PATCH v3] drivers/net: add generic ethdev macro to get PCI device
On 05/15/2017 01:24 PM, Ferruh Yigit wrote:
Instead of many PMD define their own macro, define a generic one in
ethdev and use that in PMDs.
Signed-off-by: Ferruh Yigit
Acked-by: Allain Legacy
Acked-by: Wenzhuo Lu
Acked-by: Andrew Rybchenko
---
v3:
* rebased on top of latest next-net, there were updates in ixgbe and
virtio
v2:
* Convert RTE_DEV_TO_PCI(eth_dev->device) to RTE_ETH_DEV_TO_PCI(eth_dev)
---
drivers/net/ark/ark_ethdev.c| 4 ++--
drivers/net/ark/ark_ethdev.h| 4
drivers/net/avp/avp_ethdev.c| 27 ++
drivers/net/bnx2x/bnx2x_ethdev.c| 4 ++--
drivers/net/bnxt/bnxt_ethdev.c | 6 ++---
drivers/net/bonding/rte_eth_bond_args.c | 2 +-
drivers/net/cxgbe/cxgbe_ethdev.c| 4 ++--
drivers/net/e1000/e1000_ethdev.h| 2 --
drivers/net/e1000/em_ethdev.c | 15 ++---
drivers/net/e1000/igb_ethdev.c | 28 +++
drivers/net/e1000/igb_pf.c | 2 +-
drivers/net/ena/ena_ethdev.c| 4 ++--
drivers/net/enic/enic_ethdev.c | 4 ++--
drivers/net/fm10k/fm10k_ethdev.c| 16 ++---
drivers/net/i40e/i40e_ethdev.c | 30 -
drivers/net/i40e/i40e_ethdev.h | 3 ---
drivers/net/i40e/i40e_ethdev_vf.c | 20 -
drivers/net/ixgbe/ixgbe_ethdev.c| 40 -
drivers/net/ixgbe/ixgbe_ethdev.h| 3 ---
drivers/net/ixgbe/ixgbe_pf.c| 2 +-
drivers/net/ixgbe/rte_pmd_ixgbe.c | 20 -
drivers/net/liquidio/lio_ethdev.c | 2 +-
drivers/net/mlx4/mlx4.c | 2 +-
drivers/net/mlx5/mlx5_ethdev.c | 2 +-
drivers/net/nfp/nfp_net.c | 16 ++---
drivers/net/qede/qede_ethdev.c | 6 ++---
drivers/net/sfc/sfc.c | 4 ++--
drivers/net/sfc/sfc.h | 3 ---
drivers/net/sfc/sfc_debug.h | 3 ++-
drivers/net/sfc/sfc_ethdev.c| 4 ++--
drivers/net/sfc/sfc_intr.c | 10 -
drivers/net/sfc/sfc_log.h | 3 ++-
drivers/net/sfc/sfc_rx.c| 2 +-
drivers/net/sfc/sfc_tx.c| 2 +-
drivers/net/szedata2/rte_eth_szedata2.c | 6 ++---
drivers/net/thunderx/nicvf_ethdev.c | 6 ++---
drivers/net/virtio/virtio_ethdev.c | 11 +
drivers/net/vmxnet3/vmxnet3_ethdev.c| 4 ++--
lib/librte_ether/rte_ethdev.h | 2 ++
39 files changed, 156 insertions(+), 172 deletions(-)
diff --git a/drivers/net/ark/ark_ethdev.c b/drivers/net/ark/ark_ethdev.c
index 995c93d..5c45038 100644
--- a/drivers/net/ark/ark_ethdev.c
+++ b/drivers/net/ark/ark_ethdev.c
@@ -278,7 +278,7 @@ eth_ark_dev_init(struct rte_eth_dev *dev)
ret = check_for_ext(ark);
if (ret)
return ret;
- pci_dev = ARK_DEV_TO_PCI(dev);
+ pci_dev = RTE_ETH_DEV_TO_PCI(dev);
rte_eth_copy_pci_info(dev, pci_dev);
/* Use dummy function until setup */
@@ -751,7 +751,7 @@ eth_ark_dev_info_get(struct rte_eth_dev *dev,
ETH_LINK_SPEED_40G |
ETH_LINK_SPEED_50G |
ETH_LINK_SPEED_100G);
- dev_info->pci_dev = ARK_DEV_TO_PCI(dev);
+ dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
}
static int
diff --git a/drivers/net/ark/ark_ethdev.h b/drivers/net/ark/ark_ethdev.h
index 9f8d32f..df5547b 100644
--- a/drivers/net/ark/ark_ethdev.h
+++ b/drivers/net/ark/ark_ethdev.h
@@ -34,8 +34,4 @@
#ifndef _ARK_ETHDEV_H_
#define _ARK_ETHDEV_H_
-#define ARK_DEV_TO_PCI(eth_dev) \
- RTE_DEV_TO_PCI((eth_dev)->device)
-
-
#endif
diff --git a/drivers/net/avp/avp_ethdev.c b/drivers/net/avp/avp_ethdev.c
index b08dbaa..3554645 100644
--- a/drivers/net/avp/avp_ethdev.c
+++ b/drivers/net/avp/avp_ethdev.c
@@ -112,9 +112,6 @@ static void avp_dev_stats_get(struct rte_eth_dev *dev,
static void avp_dev_stats_reset(struct rte_eth_dev *dev);
-#define AVP_DEV_TO_PCI(eth_dev) RTE_DEV_TO_PCI((eth_dev)->device)
-
-
#define AVP_MAX_RX_BURST 64
#define AVP_MAX_TX_BURST 64
#define AVP_MAX_MAC_ADDRS 1
@@ -392,7 +389,7 @@ static void *
avp_dev_translate_address(struct rte_eth_dev *eth_dev,
phys_addr_t host_phys_addr)
{
- struct rte_pci_device *pci_dev = AVP_DEV_TO_PCI(eth_dev);
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
struct rte_mem_resource *resource;
struct rte_avp_memmap_info *info;
struct rte_avp_memmap *map;
@@ -445,7 +442,7 @@ avp_dev_version_check(uint32_t version)
static int
avp_dev_check_regions(struct rte_eth_dev *eth_dev)
{
- struct rte_pci_device *pci_dev = AVP_DEV_TO_PCI(eth_dev);
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
struct rte_avp_memmap_info *memmap;
[dpdk-dev] [PATCH v3] drivers/net: add generic ethdev macro to get PCI device
Instead of many PMD define their own macro, define a generic one in
ethdev and use that in PMDs.
Signed-off-by: Ferruh Yigit
Acked-by: Allain Legacy
Acked-by: Wenzhuo Lu
---
v3:
* rebased on top of latest next-net, there were updates in ixgbe and
virtio
v2:
* Convert RTE_DEV_TO_PCI(eth_dev->device) to RTE_ETH_DEV_TO_PCI(eth_dev)
---
drivers/net/ark/ark_ethdev.c| 4 ++--
drivers/net/ark/ark_ethdev.h| 4
drivers/net/avp/avp_ethdev.c| 27 ++
drivers/net/bnx2x/bnx2x_ethdev.c| 4 ++--
drivers/net/bnxt/bnxt_ethdev.c | 6 ++---
drivers/net/bonding/rte_eth_bond_args.c | 2 +-
drivers/net/cxgbe/cxgbe_ethdev.c| 4 ++--
drivers/net/e1000/e1000_ethdev.h| 2 --
drivers/net/e1000/em_ethdev.c | 15 ++---
drivers/net/e1000/igb_ethdev.c | 28 +++
drivers/net/e1000/igb_pf.c | 2 +-
drivers/net/ena/ena_ethdev.c| 4 ++--
drivers/net/enic/enic_ethdev.c | 4 ++--
drivers/net/fm10k/fm10k_ethdev.c| 16 ++---
drivers/net/i40e/i40e_ethdev.c | 30 -
drivers/net/i40e/i40e_ethdev.h | 3 ---
drivers/net/i40e/i40e_ethdev_vf.c | 20 -
drivers/net/ixgbe/ixgbe_ethdev.c| 40 -
drivers/net/ixgbe/ixgbe_ethdev.h| 3 ---
drivers/net/ixgbe/ixgbe_pf.c| 2 +-
drivers/net/ixgbe/rte_pmd_ixgbe.c | 20 -
drivers/net/liquidio/lio_ethdev.c | 2 +-
drivers/net/mlx4/mlx4.c | 2 +-
drivers/net/mlx5/mlx5_ethdev.c | 2 +-
drivers/net/nfp/nfp_net.c | 16 ++---
drivers/net/qede/qede_ethdev.c | 6 ++---
drivers/net/sfc/sfc.c | 4 ++--
drivers/net/sfc/sfc.h | 3 ---
drivers/net/sfc/sfc_debug.h | 3 ++-
drivers/net/sfc/sfc_ethdev.c| 4 ++--
drivers/net/sfc/sfc_intr.c | 10 -
drivers/net/sfc/sfc_log.h | 3 ++-
drivers/net/sfc/sfc_rx.c| 2 +-
drivers/net/sfc/sfc_tx.c| 2 +-
drivers/net/szedata2/rte_eth_szedata2.c | 6 ++---
drivers/net/thunderx/nicvf_ethdev.c | 6 ++---
drivers/net/virtio/virtio_ethdev.c | 11 +
drivers/net/vmxnet3/vmxnet3_ethdev.c| 4 ++--
lib/librte_ether/rte_ethdev.h | 2 ++
39 files changed, 156 insertions(+), 172 deletions(-)
diff --git a/drivers/net/ark/ark_ethdev.c b/drivers/net/ark/ark_ethdev.c
index 995c93d..5c45038 100644
--- a/drivers/net/ark/ark_ethdev.c
+++ b/drivers/net/ark/ark_ethdev.c
@@ -278,7 +278,7 @@ eth_ark_dev_init(struct rte_eth_dev *dev)
ret = check_for_ext(ark);
if (ret)
return ret;
- pci_dev = ARK_DEV_TO_PCI(dev);
+ pci_dev = RTE_ETH_DEV_TO_PCI(dev);
rte_eth_copy_pci_info(dev, pci_dev);
/* Use dummy function until setup */
@@ -751,7 +751,7 @@ eth_ark_dev_info_get(struct rte_eth_dev *dev,
ETH_LINK_SPEED_40G |
ETH_LINK_SPEED_50G |
ETH_LINK_SPEED_100G);
- dev_info->pci_dev = ARK_DEV_TO_PCI(dev);
+ dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
}
static int
diff --git a/drivers/net/ark/ark_ethdev.h b/drivers/net/ark/ark_ethdev.h
index 9f8d32f..df5547b 100644
--- a/drivers/net/ark/ark_ethdev.h
+++ b/drivers/net/ark/ark_ethdev.h
@@ -34,8 +34,4 @@
#ifndef _ARK_ETHDEV_H_
#define _ARK_ETHDEV_H_
-#define ARK_DEV_TO_PCI(eth_dev)\
- RTE_DEV_TO_PCI((eth_dev)->device)
-
-
#endif
diff --git a/drivers/net/avp/avp_ethdev.c b/drivers/net/avp/avp_ethdev.c
index b08dbaa..3554645 100644
--- a/drivers/net/avp/avp_ethdev.c
+++ b/drivers/net/avp/avp_ethdev.c
@@ -112,9 +112,6 @@ static void avp_dev_stats_get(struct rte_eth_dev *dev,
static void avp_dev_stats_reset(struct rte_eth_dev *dev);
-#define AVP_DEV_TO_PCI(eth_dev) RTE_DEV_TO_PCI((eth_dev)->device)
-
-
#define AVP_MAX_RX_BURST 64
#define AVP_MAX_TX_BURST 64
#define AVP_MAX_MAC_ADDRS 1
@@ -392,7 +389,7 @@ static void *
avp_dev_translate_address(struct rte_eth_dev *eth_dev,
phys_addr_t host_phys_addr)
{
- struct rte_pci_device *pci_dev = AVP_DEV_TO_PCI(eth_dev);
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
struct rte_mem_resource *resource;
struct rte_avp_memmap_info *info;
struct rte_avp_memmap *map;
@@ -445,7 +442,7 @@ avp_dev_version_check(uint32_t version)
static int
avp_dev_check_regions(struct rte_eth_dev *eth_dev)
{
- struct rte_pci_device *pci_dev = AVP_DEV_TO_PCI(eth_dev);
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
struct rte_avp_memmap_info *memmap;
struct rte_avp_device_info *info;
struct rte_mem_resource *resource;
@@ -581,7 +578,7 @@ _avp_set_rx_q
Re: [dpdk-dev] SIMD Rx/Tx paths
On Mon, May 15, 2017 at 02:35:55PM +0200, Thomas Monjalon wrote: > Hi, > > I would like to open a discussion about SIMD code in drivers. > > I think we should not have different behaviours or features capabilities, > in the different code paths of a same driver. > I suggest to consider such differences as exceptions. > So we should merge features files (i.e. matrix columns), > and remove these files: > > % ls doc/guides/nics/features/*_vec.ini > > doc/guides/nics/features/fm10k_vec.ini > doc/guides/nics/features/fm10k_vf_vec.ini > doc/guides/nics/features/i40e_vec.ini > doc/guides/nics/features/i40e_vf_vec.ini > doc/guides/nics/features/ixgbe_vec.ini > doc/guides/nics/features/ixgbe_vf_vec.ini > doc/guides/nics/features/virtio_vec.ini > > If a feature is not supported in all code paths of a driver, > it must be marked as partially (P) supported. > > Then the mid-term goal will be to try removing these inconsistencies. > > Opinions/comments? Yes, there are inconsistencies, but if they are hidden from the user, e.g. by having the driver select automatically the most appropriate path, I don't think we should need to mark the support as "partial". Instead, it should be marked as being fully supported, but perhaps with a note indicating that a performance hit may be experienced due to the code taking a less-optimised driver path. After all, especially in the TX code path, a lot of the speed-up comes from not supporting different features, as well as from the vectorization. In those cases "closing the gap" may mean losing performance for those who don't want the features, which is not acceptable. Any feature support we can add, without affecting performance, should of course be implemented. /Bruce
Re: [dpdk-dev] SIMD Rx/Tx paths
On 5/15/2017 2:15 PM, Bruce Richardson wrote: > On Mon, May 15, 2017 at 02:35:55PM +0200, Thomas Monjalon wrote: >> Hi, >> >> I would like to open a discussion about SIMD code in drivers. >> >> I think we should not have different behaviours or features capabilities, >> in the different code paths of a same driver. >> I suggest to consider such differences as exceptions. >> So we should merge features files (i.e. matrix columns), >> and remove these files: >> >> % ls doc/guides/nics/features/*_vec.ini >> >> doc/guides/nics/features/fm10k_vec.ini >> doc/guides/nics/features/fm10k_vf_vec.ini >> doc/guides/nics/features/i40e_vec.ini >> doc/guides/nics/features/i40e_vf_vec.ini >> doc/guides/nics/features/ixgbe_vec.ini >> doc/guides/nics/features/ixgbe_vf_vec.ini >> doc/guides/nics/features/virtio_vec.ini >> >> If a feature is not supported in all code paths of a driver, >> it must be marked as partially (P) supported. >> >> Then the mid-term goal will be to try removing these inconsistencies. >> >> Opinions/comments? > > Yes, there are inconsistencies, but if they are hidden from the user, > e.g. by having the driver select automatically the most appropriate > path, I don't think we should need to mark the support as "partial". > Instead, it should be marked as being fully supported, but perhaps with > a note indicating that a performance hit may be experienced due to the > code taking a less-optimised driver path. After all, especially in the > TX code path, a lot of the speed-up comes from not supporting different > features, as well as from the vectorization. In those cases "closing the > gap" may mean losing performance for those who don't want the features, > which is not acceptable. Any feature support we can add, without > affecting performance, should of course be implemented. I mostly agree with Bruce, except for PMD selection the patch automatically. There is a trade off between feature set and performance, scalar driver favors features and vector driver favors performance, I think good to have both. And I agree that feature support should be added to vector drivers as long as it doesn't effect the performance. Related to the driver auto selecting the path, I concern this may confuse the user, because he may end up a situation he doesn't clear about supported features, I am for more explicit way to select the scalar or vector driver. And for merging the features files, most of the items are already same for scalar and vector drivers, so perhaps we can merge files and use different syntax for features that is different for scalar and vector: Ys: Yes Scalar [no vector] Yv: Yes Vector [no scalar] Y: Yes Both Ps: Partially Scalar [no vector] Pv: Partially Vector [no scalar] P: Partially Both YsPv, YvPs > > /Bruce >
Re: [dpdk-dev] SIMD Rx/Tx paths
> -Original Message- > From: Yigit, Ferruh > Sent: Monday, May 15, 2017 2:36 PM > To: Richardson, Bruce ; Thomas Monjalon > > Cc: dev@dpdk.org; Yuanhan Liu ; Maxime > Coquelin ; Chen, Jing D > ; Zhang, Helin ; Wu, > Jingjing ; Lu, Wenzhuo ; > Ananyev, Konstantin > Subject: Re: SIMD Rx/Tx paths > > On 5/15/2017 2:15 PM, Bruce Richardson wrote: > > On Mon, May 15, 2017 at 02:35:55PM +0200, Thomas Monjalon wrote: > >> Hi, > >> > >> I would like to open a discussion about SIMD code in drivers. > >> > >> I think we should not have different behaviours or features > >> capabilities, in the different code paths of a same driver. > >> I suggest to consider such differences as exceptions. > >> So we should merge features files (i.e. matrix columns), and remove > >> these files: > >> > >> % ls doc/guides/nics/features/*_vec.ini > >> > >> doc/guides/nics/features/fm10k_vec.ini > >> doc/guides/nics/features/fm10k_vf_vec.ini > >> doc/guides/nics/features/i40e_vec.ini > >> doc/guides/nics/features/i40e_vf_vec.ini > >> doc/guides/nics/features/ixgbe_vec.ini > >> doc/guides/nics/features/ixgbe_vf_vec.ini > >> doc/guides/nics/features/virtio_vec.ini > >> > >> If a feature is not supported in all code paths of a driver, it must > >> be marked as partially (P) supported. > >> > >> Then the mid-term goal will be to try removing these inconsistencies. > >> > >> Opinions/comments? > > > > Yes, there are inconsistencies, but if they are hidden from the user, > > e.g. by having the driver select automatically the most appropriate > > path, I don't think we should need to mark the support as "partial". > > Instead, it should be marked as being fully supported, but perhaps > > with a note indicating that a performance hit may be experienced due > > to the code taking a less-optimised driver path. After all, especially > > in the TX code path, a lot of the speed-up comes from not supporting > > different features, as well as from the vectorization. In those cases > > "closing the gap" may mean losing performance for those who don't want > > the features, which is not acceptable. Any feature support we can add, > > without affecting performance, should of course be implemented. > > I mostly agree with Bruce, except for PMD selection the patch > automatically. > > There is a trade off between feature set and performance, scalar driver > favors features and vector driver favors performance, I think good to have > both. > > And I agree that feature support should be added to vector drivers as long > as it doesn't effect the performance. > > Related to the driver auto selecting the path, I concern this may confuse > the user, because he may end up a situation he doesn't clear about > supported features, I am for more explicit way to select the scalar or > vector driver. > > And for merging the features files, most of the items are already same for > scalar and vector drivers, so perhaps we can merge files and use different > syntax for features that is different for scalar and vector: > Ys: Yes Scalar [no vector] > Yv: Yes Vector [no scalar] > Y: Yes Both > Ps: Partially Scalar [no vector] > Pv: Partially Vector [no scalar] > P: Partially Both > YsPv, YvPs > For the table, I don't really mind so much what scheme is agreed. For the user doing the coding, while I can accept that it might be useful to support explicitly request a vector or scalar driver, I'd definitely prefer the default state to remain auto-selection based on features requested. If a user want TSO, then pick the best driver path that supports TSO, and don't force the user to read up on what the different paths are! /Bruce
Re: [dpdk-dev] SIMD Rx/Tx paths
15/05/2017 16:12, Richardson, Bruce: > From: Yigit, Ferruh > > On 5/15/2017 2:15 PM, Bruce Richardson wrote: > > > On Mon, May 15, 2017 at 02:35:55PM +0200, Thomas Monjalon wrote: > > >> Hi, > > >> > > >> I would like to open a discussion about SIMD code in drivers. > > >> > > >> I think we should not have different behaviours or features > > >> capabilities, in the different code paths of a same driver. > > >> I suggest to consider such differences as exceptions. > > >> So we should merge features files (i.e. matrix columns), and remove > > >> these files: > > >> > > >> % ls doc/guides/nics/features/*_vec.ini > > >> > > >> doc/guides/nics/features/fm10k_vec.ini > > >> doc/guides/nics/features/fm10k_vf_vec.ini > > >> doc/guides/nics/features/i40e_vec.ini > > >> doc/guides/nics/features/i40e_vf_vec.ini > > >> doc/guides/nics/features/ixgbe_vec.ini > > >> doc/guides/nics/features/ixgbe_vf_vec.ini > > >> doc/guides/nics/features/virtio_vec.ini > > >> > > >> If a feature is not supported in all code paths of a driver, it must > > >> be marked as partially (P) supported. > > >> > > >> Then the mid-term goal will be to try removing these inconsistencies. > > >> > > >> Opinions/comments? > > > > > > Yes, there are inconsistencies, but if they are hidden from the user, > > > e.g. by having the driver select automatically the most appropriate > > > path, I don't think we should need to mark the support as "partial". > > > Instead, it should be marked as being fully supported, but perhaps > > > with a note indicating that a performance hit may be experienced due > > > to the code taking a less-optimised driver path. After all, especially > > > in the TX code path, a lot of the speed-up comes from not supporting > > > different features, as well as from the vectorization. In those cases > > > "closing the gap" may mean losing performance for those who don't want > > > the features, which is not acceptable. Any feature support we can add, > > > without affecting performance, should of course be implemented. > > > > I mostly agree with Bruce, except for PMD selection the patch > > automatically. > > > > There is a trade off between feature set and performance, scalar driver > > favors features and vector driver favors performance, I think good to have > > both. > > > > And I agree that feature support should be added to vector drivers as long > > as it doesn't effect the performance. > > > > Related to the driver auto selecting the path, I concern this may confuse > > the user, because he may end up a situation he doesn't clear about > > supported features, I am for more explicit way to select the scalar or > > vector driver. > > > > And for merging the features files, most of the items are already same for > > scalar and vector drivers, so perhaps we can merge files and use different > > syntax for features that is different for scalar and vector: > > Ys: Yes Scalar [no vector] > > Yv: Yes Vector [no scalar] > > Y: Yes Both > > Ps: Partially Scalar [no vector] > > Pv: Partially Vector [no scalar] > > P: Partially Both > > YsPv, YvPs Please remember that there are different vector code paths (SSE/AVX, NEON, Altivec). > For the table, I don't really mind so much what scheme is agreed. For the > user doing the coding, while I can accept that it might be useful to support > explicitly request a vector or scalar driver, I'd definitely prefer the > default state to remain auto-selection based on features requested. If a user > want TSO, then pick the best driver path that supports TSO, and don't force > the user to read up on what the different paths are! I agree. If we can be sure what the application needs, we can select the best code path and mark the feature supported. But can we be sure of the expectations for every features? How do we know that the application relies on certain packet types (which are not recognized by some code paths)?
[dpdk-dev] [RFC] doc: document NIC features
Document NIC features, add more information about them and add more implementation related support. TODO: - Not all features all fully documented, need help from community - Instead of having a new file, it would be nice to auto generate this file from defaults.ini, and perhaps add extra information as comment to that file. - Some features are implementing eth_dev_ops and some are updates in Rx/Tx path. These features can be grouped together. - A link from overview file per feature would be nice. Signed-off-by: Ferruh Yigit --- doc/guides/nics/overview_features.rst | 396 ++ 1 file changed, 396 insertions(+) create mode 100644 doc/guides/nics/overview_features.rst diff --git a/doc/guides/nics/overview_features.rst b/doc/guides/nics/overview_features.rst new file mode 100644 index 000..26c83b6 --- /dev/null +++ b/doc/guides/nics/overview_features.rst @@ -0,0 +1,396 @@ +.. BSD LICENSE +Copyright(c) 2017 Intel Corporation. All rights reserved. +All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: + +* Redistributions of source code must retain the above copyright +notice, this list of conditions and the following disclaimer. +* Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in +the documentation and/or other materials provided with the +distribution. +* Neither the name of Intel Corporation nor the names of its +contributors may be used to endorse or promote products derived +from this software without specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +Overview features += + +Speed capabilities +-- +eth_dev_ops: dev_infos_get:speed_capa + +Driver advertise the speed capabilities that the current device capable to offer. + + +Link status +--- +eth_dev_ops: link_update + + +Link status event +- +rte_pci_driver.drv_flags = RTE_PCI_DRV_INTR_LSC + + +Removal event +- +rte_pci_driver.drv_flags = RTE_PCI_DRV_INTR_RMV + + +Queue status event +-- +eth_dev_ops: link_update + + +Rx interrupt + +eth_dev_ops: rx_queue_intr_enable, rx_queue_intr_disable + + +Free Tx mbuf on demand +-- +eth_dev_ops: tx_done_cleanup + + +Queue start/stop + +eth_dev_ops: rx_queue_start, rx_queue_stop, tx_queue_start, tx_queue_stop + + +MTU update +-- +eth_dev_ops: mtu_set + + +Jumbo frame +--- + + +Scattered Rx + + + +LRO +--- +Large Receive Offload. + + +TSO +--- +TCP Segmentation Offloading. + + +Promiscuous mode + +eth_dev_ops: promiscuous_enable, promiscuous_disable + + +Allmulticast mode +- +eth_dev_ops: allmulticast_enable, allmulticast_disable + + +Unicast MAC filter +-- +eth_dev_ops: mac_addr_add, mac_addr_remove + + +Multicast MAC filter + +eth_dev_ops: set_mc_addr_list + + +RSS hash + +eth_dev_ops: rss_hash_update, rss_hash_conf_get + + +RSS key update +-- +eth_dev_ops: rss_hash_update, rss_hash_conf_get + + +RSS reta update +--- +eth_dev_ops: reta_update, reta_query + + +VMDq + + + +SR-IOV +-- + + +DCB +--- +eth_dev_ops: get_dcb_info + + +VLAN filter +--- +eth_dev_ops: vlan_filter_set + + +Ethertype filter + +eth_dev_ops: filter_ctrl:RTE_ETH_FILTER_ETHERTYPE + + +N-tuple filter +-- +eth_dev_ops: filter_ctrl:RTE_ETH_FILTER_NTUPLE + + +SYN filter +-- +eth_dev_ops: filter_ctrl:RTE_ETH_FILTER_SYN + + +Tunnel filter +- +eth_dev_ops: filter_ctrl:RTE_ETH_FILTER_TUNNEL + + +Flexible filter +--- +eth_dev_ops: filter_ctrl:RTE_ETH_FILTER_FLEXIBLE + + +Hash filter +--- +eth_dev_ops: filter_ctrl:RTE_ETH_FILTER_HASH + + +Flow director +- +eth_dev_ops: filter_ctrl:RTE_ETH_FILTER_FDIR + + +Flow control + +eth_dev_ops: flow_ctrl_get, flow_ctrl_set, priority_flow_ctrl_set + + +Flow API + +eth_dev_ops:
Re: [dpdk-dev] [PATCH v2 01/20] mk: change to cortex-a72
On 5/15/2017 1:37 PM, Hemant Agrawal wrote: > Signed-off-by: Hemant Agrawal > --- > config/defconfig_arm64-dpaa2-linuxapp-gcc | 2 +- > mk/machine/dpaa2/rte.vars.mk | 2 +- > 2 files changed, 2 insertions(+), 2 deletions(-) > > diff --git a/config/defconfig_arm64-dpaa2-linuxapp-gcc > b/config/defconfig_arm64-dpaa2-linuxapp-gcc > index 314a0ec..2304ab6 100644 > --- a/config/defconfig_arm64-dpaa2-linuxapp-gcc > +++ b/config/defconfig_arm64-dpaa2-linuxapp-gcc > @@ -34,7 +34,7 @@ > > # NXP (Freescale) - Soc Architecture with WRIOP and QBMAN support > CONFIG_RTE_MACHINE="dpaa2" > -CONFIG_RTE_ARCH_ARM_TUNE="cortex-a57+fp+simd" > +CONFIG_RTE_ARCH_ARM_TUNE="cortex-a72" Can you please describe reasoning and effect in commit log: What will happen when cortex-a72 used instead of cortex-a57+fp+simd? And what is the motivation for change? Thanks, ferruh
Re: [dpdk-dev] [PATCH v2 02/20] net/dpaa2: remove port level buffer layout definition
On 5/15/2017 1:37 PM, Hemant Agrawal wrote: > Signed-off-by: Hemant Agrawal Same comment here, can you please mention in the commit log why this piece of code removed, what will be effect of the removal. Thanks, ferruh
Re: [dpdk-dev] [PATCH v2 07/20] bus/fslmc: export qbman results in map file
On 5/15/2017 1:37 PM, Hemant Agrawal wrote:
> Signed-off-by: Hemant Agrawal
> ---
> drivers/bus/fslmc/rte_bus_fslmc_version.map | 8
> 1 file changed, 8 insertions(+)
>
> diff --git a/drivers/bus/fslmc/rte_bus_fslmc_version.map
> b/drivers/bus/fslmc/rte_bus_fslmc_version.map
> index 2db0fce..6e7b94d 100644
> --- a/drivers/bus/fslmc/rte_bus_fslmc_version.map
> +++ b/drivers/bus/fslmc/rte_bus_fslmc_version.map
> @@ -49,3 +49,11 @@ DPDK_17.05 {
>
> local: *;
> };
> +
> +DPDK_17.08 {
> + global:
> +
> + qbman_result_SCN_state_in_mem;
> +
> + local: *;
> +};
Syntax should be:
DPDK_17.08 {
global:
qbman_result_SCN_state_in_mem;
} DPDK_17.05;
No "local: *", and previous group (DPDK_17.05) after "}".
But again, it would be nice if you can document in commit log, why
exporting this variable required now?
Thanks,
ferruh
[dpdk-dev] [PATCH v2] eventdev: clarify atomic and ordered queue config
The nb_atomic_flows and nb_atomic_order_sequences fields are only inspected
if the queue is configured for atomic or ordered scheduling, respectively.
This commit updates the documentation to reflect that.
Signed-off-by: Gage Eads
---
v2: Fixed doxygen output issue and tweaked the ranges
lib/librte_eventdev/rte_eventdev.h | 15 ++-
1 file changed, 10 insertions(+), 5 deletions(-)
diff --git a/lib/librte_eventdev/rte_eventdev.h
b/lib/librte_eventdev/rte_eventdev.h
index 20e7293..9428433 100644
--- a/lib/librte_eventdev/rte_eventdev.h
+++ b/lib/librte_eventdev/rte_eventdev.h
@@ -521,9 +521,11 @@ rte_event_dev_configure(uint8_t dev_id,
struct rte_event_queue_conf {
uint32_t nb_atomic_flows;
/**< The maximum number of active flows this queue can track at any
-* given time. The value must be in the range of
-* [1 - nb_event_queue_flows)] which previously provided in
-* rte_event_dev_info_get().
+* given time. If the queue is configured for atomic scheduling (by
+* applying the RTE_EVENT_QUEUE_CFG_ALL_TYPES or
+* RTE_EVENT_QUEUE_CFG_ATOMIC_ONLY flags to event_queue_cfg), then the
+* value must be in the range of [1, nb_event_queue_flows], which was
+* previously provided in rte_event_dev_configure().
*/
uint32_t nb_atomic_order_sequences;
/**< The maximum number of outstanding events waiting to be
@@ -533,8 +535,11 @@ struct rte_event_queue_conf {
* scheduler cannot schedule the events from this queue and invalid
* event will be returned from dequeue until one or more entries are
* freed up/released.
-* The value must be in the range of [1 - nb_event_queue_flows)]
-* which previously supplied to rte_event_dev_configure().
+* If the queue is configured for ordered scheduling (by applying the
+* RTE_EVENT_QUEUE_CFG_ALL_TYPES or RTE_EVENT_QUEUE_CFG_ORDERED_ONLY
+* flags to event_queue_cfg), then the value must be in the range of
+* [1, nb_event_queue_flows], which was previously supplied to
+* rte_event_dev_configure().
*/
uint32_t event_queue_cfg; /**< Queue cfg flags(EVENT_QUEUE_CFG_) */
uint8_t priority;
--
2.7.4
Re: [dpdk-dev] SIMD Rx/Tx paths
> 15/05/2017 16:12, Richardson, Bruce: > > From: Yigit, Ferruh > > > On 5/15/2017 2:15 PM, Bruce Richardson wrote: > > > > On Mon, May 15, 2017 at 02:35:55PM +0200, Thomas Monjalon wrote: > > > >> Hi, > > > >> > > > >> I would like to open a discussion about SIMD code in drivers. > > > >> > > > >> I think we should not have different behaviours or features > > > >> capabilities, in the different code paths of a same driver. > > > >> I suggest to consider such differences as exceptions. > > > >> So we should merge features files (i.e. matrix columns), and > > > >> remove these files: > > > >> > > > >> % ls doc/guides/nics/features/*_vec.ini > > > >> > > > >> doc/guides/nics/features/fm10k_vec.ini > > > >> doc/guides/nics/features/fm10k_vf_vec.ini > > > >> doc/guides/nics/features/i40e_vec.ini > > > >> doc/guides/nics/features/i40e_vf_vec.ini > > > >> doc/guides/nics/features/ixgbe_vec.ini > > > >> doc/guides/nics/features/ixgbe_vf_vec.ini > > > >> doc/guides/nics/features/virtio_vec.ini > > > >> > > > >> If a feature is not supported in all code paths of a driver, it > > > >> must be marked as partially (P) supported. > > > >> > > > >> Then the mid-term goal will be to try removing these inconsistencies. > > > >> > > > >> Opinions/comments? > > > > > > > > Yes, there are inconsistencies, but if they are hidden from the > > > > user, e.g. by having the driver select automatically the most > > > > appropriate path, I don't think we should need to mark the support as > "partial". > > > > Instead, it should be marked as being fully supported, but perhaps > > > > with a note indicating that a performance hit may be experienced > > > > due to the code taking a less-optimised driver path. After all, > > > > especially in the TX code path, a lot of the speed-up comes from > > > > not supporting different features, as well as from the > > > > vectorization. In those cases "closing the gap" may mean losing > > > > performance for those who don't want the features, which is not > > > > acceptable. Any feature support we can add, without affecting > performance, should of course be implemented. > > > > > > I mostly agree with Bruce, except for PMD selection the patch > > > automatically. > > > > > > There is a trade off between feature set and performance, scalar > > > driver favors features and vector driver favors performance, I think > > > good to have both. > > > > > > And I agree that feature support should be added to vector drivers > > > as long as it doesn't effect the performance. > > > > > > Related to the driver auto selecting the path, I concern this may > > > confuse the user, because he may end up a situation he doesn't clear > > > about supported features, I am for more explicit way to select the > > > scalar or vector driver. > > > > > > And for merging the features files, most of the items are already > > > same for scalar and vector drivers, so perhaps we can merge files > > > and use different syntax for features that is different for scalar and > > > vector: > > > Ys: Yes Scalar [no vector] > > > Yv: Yes Vector [no scalar] > > > Y: Yes Both > > > Ps: Partially Scalar [no vector] > > > Pv: Partially Vector [no scalar] > > > P: Partially Both > > > YsPv, YvPs > > Please remember that there are different vector code paths (SSE/AVX, NEON, > Altivec). > > > For the table, I don't really mind so much what scheme is agreed. For the > user doing the coding, while I can accept that it might be useful to support > explicitly request a vector or scalar driver, I'd definitely prefer the > default > state to remain auto-selection based on features requested. If a user want > TSO, then pick the best driver path that supports TSO, and don't force the > user to read up on what the different paths are! > > I agree. > If we can be sure what the application needs, we can select the best code > path and mark the feature supported. > But can we be sure of the expectations for every features? > How do we know that the application relies on certain packet types (which > are not recognized by some code paths)? I also agree auto-selection on tx/rx func. User needn't worry about how PMD to satisfy its' requirement, result is more important. Besides that, we should do more work in rx/tx configuration to help PMD better decide the best rx/tx. Pkt_type is a good example. A possible way is to expose all possible PMD offload features into structure rte_eth_rxmode and rte_eth_txmode or a new structure.
Re: [dpdk-dev] [PATCH v2] test: add delay time in test alarm
Hi,
> -Original Message-
> From: Chen, Jing D
> Sent: Friday, May 5, 2017 2:29 PM
> To: Yang, Qiming ; dev@dpdk.org
> Cc: Wu, Jingjing
> Subject: RE: [PATCH v2] test: add delay time in test alarm
>
> Hi,
>
> > diff --git a/test/test/test_alarm.c b/test/test/test_alarm.c index
> > ecb2f6d..cbae1a0 100644
> > --- a/test/test/test_alarm.c
> > +++ b/test/test/test_alarm.c
> > @@ -96,6 +96,7 @@ static int
> > test_multi_alarms(void)
> > {
> > int rm_count = 0;
> > + int count = 0;
> > cb_count.cnt = 0;
> >
> > printf("Expect 6 callbacks in order...\n"); @@ -169,7 +170,10 @@
> > test_multi_alarms(void)
> > printf("Error, cancelling head-of-list leads to premature
> > callback\n");
> > return -1;
> > }
> > - rte_delay_ms(10);
> > +
> > + while (flag != 2 && count++ < 6)
> > + rte_delay_ms(10);
The count can be replaced with macro, but the delay 10 ms is based on the alarm
time set before.
rte_eal_alarm_set(10 * US_PER_MS, test_remove_in_callback, (void *)1);
> > +
> > if (flag != 2) {
> > printf("Error - expected callback not called\n");
> > rte_eal_alarm_cancel(test_remove_in_callback, (void *)-1);
> @@
> > -212,7 +216,7 @@ test_alarm(void)
> > printf("fail to set alarm callback\n");
> > return -1;
> > }
> > - while (flag == 0 && count ++ < 6)
> > + while (flag == 0 && count++ < 20)
> > rte_delay_ms(RTE_TEST_CHECK_PERIOD);
> >
>
> What's the criteria to delay 20* RTE_TEST_CHECK_PERIOD ms? Add more
> comments?
>
> > if (flag == 0){
> > --
> > 2.7.4
>
> Overall comment is to replace numeric with macro.
Re: [dpdk-dev] [PATCH v2] eventdev: clarify atomic and ordered queue config
-Original Message-
> Date: Mon, 15 May 2017 14:40:14 -0500
> From: Gage Eads
> To: dev@dpdk.org
> CC: jerin.ja...@caviumnetworks.com
> Subject: [PATCH v2] eventdev: clarify atomic and ordered queue config
> X-Mailer: git-send-email 2.7.4
>
> The nb_atomic_flows and nb_atomic_order_sequences fields are only inspected
> if the queue is configured for atomic or ordered scheduling, respectively.
> This commit updates the documentation to reflect that.
>
> Signed-off-by: Gage Eads
Acked-by: Jerin Jacob
> ---
> v2: Fixed doxygen output issue and tweaked the ranges
>
> lib/librte_eventdev/rte_eventdev.h | 15 ++-
> 1 file changed, 10 insertions(+), 5 deletions(-)
>
> diff --git a/lib/librte_eventdev/rte_eventdev.h
> b/lib/librte_eventdev/rte_eventdev.h
> index 20e7293..9428433 100644
> --- a/lib/librte_eventdev/rte_eventdev.h
> +++ b/lib/librte_eventdev/rte_eventdev.h
> @@ -521,9 +521,11 @@ rte_event_dev_configure(uint8_t dev_id,
> struct rte_event_queue_conf {
> uint32_t nb_atomic_flows;
> /**< The maximum number of active flows this queue can track at any
> - * given time. The value must be in the range of
> - * [1 - nb_event_queue_flows)] which previously provided in
> - * rte_event_dev_info_get().
> + * given time. If the queue is configured for atomic scheduling (by
> + * applying the RTE_EVENT_QUEUE_CFG_ALL_TYPES or
> + * RTE_EVENT_QUEUE_CFG_ATOMIC_ONLY flags to event_queue_cfg), then the
> + * value must be in the range of [1, nb_event_queue_flows], which was
> + * previously provided in rte_event_dev_configure().
>*/
> uint32_t nb_atomic_order_sequences;
> /**< The maximum number of outstanding events waiting to be
> @@ -533,8 +535,11 @@ struct rte_event_queue_conf {
>* scheduler cannot schedule the events from this queue and invalid
>* event will be returned from dequeue until one or more entries are
>* freed up/released.
> - * The value must be in the range of [1 - nb_event_queue_flows)]
> - * which previously supplied to rte_event_dev_configure().
> + * If the queue is configured for ordered scheduling (by applying the
> + * RTE_EVENT_QUEUE_CFG_ALL_TYPES or RTE_EVENT_QUEUE_CFG_ORDERED_ONLY
> + * flags to event_queue_cfg), then the value must be in the range of
> + * [1, nb_event_queue_flows], which was previously supplied to
> + * rte_event_dev_configure().
>*/
> uint32_t event_queue_cfg; /**< Queue cfg flags(EVENT_QUEUE_CFG_) */
> uint8_t priority;
> --
> 2.7.4
>
Re: [dpdk-dev] [PATCH] net/ixgbe: support detection of hot swapped SFP/SFP+
Hi, Srini There is a bit confusion. Your patch shows that your code is added into the function eth_ixgbevf_pci_remove( ). But it is not. It is added into the fucntion ixgbe_dev_start( ), right ? So would you please rebase it to R 17.05 ? Which type of ixgbe device id did you tested ? There are many MAC types with different device id. The function ixgbe_pf_reset_hw(hw) is called before your adding code. ixgbe_pf_reset_hw() calls hw->mac.ops.reset_hw( ) which may points to following different function for different MAC type. Ixgbe_reset_hw_82598( ) calls hw->phy.ops.init(hw) if hw->phy.reset_disable == false . Ixgbe_reset_hw_82599( ) calls hw->phy.ops.init(hw) unconditionally. ixgbe_reset_hw_X540( ) doesn't' call pw->phy.ops.init(hw). For X540, hw->phy.ops.init points to ixgbe_init_phy_ops_generic() which only initialize some function pointers. Ixgbe_rest_hw_x550em() calls hw->phy.ops.init(hw) unconditionally. And for VF, ixgbe_reset_hw_vf( ) and ixgbevf_hv_reset_hw_vf( ) don't call hw->phy.ops.init(hw) anywhere. Thanks & Best Regards -Wei > -Original Message- > From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Thomas Monjalon > Sent: Sunday, May 7, 2017 6:36 AM > To: Srinivasan J > Cc: dev@dpdk.org; Lu, Wenzhuo ; Ananyev, > Konstantin > Subject: Re: [dpdk-dev] [PATCH] net/ixgbe: support detection of hot swapped > SFP/SFP+ > > 06/05/2017 15:51, Srinivasan J: > > Hi, > >Do we need an explicit "Acked-by" keyword for this > > patch to be accepted and applied? > > Yes, given it is not a trivial patch, an ack from the maintainer is required. > Anyway, it has been submitted too late for 17.05 testing.
Re: [dpdk-dev] [PATCH v5 19/26] app/testpmd: add item raw to flow command
Hi, Adrien Mazarguil > -Original Message- > From: Adrien Mazarguil [mailto:adrien.mazarg...@6wind.com] > Sent: Friday, May 12, 2017 5:13 PM > To: Zhao1, Wei > Cc: dev@dpdk.org; Xing, Beilei ; Lu, Wenzhuo > > Subject: Re: [dpdk-dev] [PATCH v5 19/26] app/testpmd: add item raw to flow > command > > Hi Wei, > > On Thu, May 11, 2017 at 06:53:52AM +, Zhao1, Wei wrote: > > Hi, Adrien > > > > > -Original Message- > > > From: dev [mailto:dev-boun...@dpdk.org] On Behalf Of Adrien > > > Mazarguil > > > Sent: Wednesday, December 21, 2016 10:52 PM > > > To: dev@dpdk.org > > > Subject: [dpdk-dev] [PATCH v5 19/26] app/testpmd: add item raw to > > > flow command > > > > > > Matches arbitrary byte strings with properties: > > > > > > - relative: look for pattern after the previous item. > > > - search: search pattern from offset (see also limit). > > > - offset: absolute or relative offset for pattern. > > > - limit: search area limit for start of pattern. > > > - length: pattern length. > > > - pattern: byte string to look for. > > > > > > Signed-off-by: Adrien Mazarguil > > > Acked-by: Olga Shern > [...] > > #define ITEM_RAW_PATTERN_SIZE 36 > > > > The size of NIC i350 flex byte filter can accommodate the max length > > size of 128 byte, and the reason to Define it as 36 is ?If it is the max > > length > of pattern, maybe 128 is more appropriate? > > Maybe I have not understand your purpose. > > > > Thank you. > > It's more or less an arbitrary compromise due to various limitations. > > Once parsed, the result of an entire command is stored in a fixed buffer of > size CMDLINE_PARSE_RESULT_BUFSIZE (8192). Each parsed token ends up > somewhere in that buffer. > > Each flow item always consumes sizeof(struct rte_flow_item) + sizeof(struct > rte_flow_item_xxx) * 3 (spec, last and mask) + alignment constraints. > > For the raw item, this makes at least: > > (sizeof(rte_flow_item) + > (sizeof(rte_flow_item_raw) + ITEM_RAW_PATTERN_SIZE) * 3) > /* (32 + (12 + 36) * 3) => 176 bytes */ > > Because space is always consumed regardless of the size of the byte string to > match for implementation reasons, there is a chance to fill the buffer too > quickly with a larger ITEM_RAW_PATTERN_SIZE. > > Also, this does not prevent users from specifying larger raw patterns (even > larger than 128) by combining them, e.g.: > > flow create 0 > pattern eth / raw relative is 1 pattern is foobar / >raw relative is 1 pattern is barbaz / end > actions queue index 42 / end > > Such a pattern ends up matching a single "foobarbarbaz" string. > > To summarize, it is only due to testpmd limitations. Even without PMD > support for combination, the current ability to provide 36 bytes of raw data > to match per specified item is plenty to validate basic functionality. We'll > improve testpmd eventually. > Thank you for your detailed explanation. Igb flex byte filter will support for that type combination for raw item. But this testpmd limitation will make trouble for users and tester. > -- > Adrien Mazarguil > 6WIND
Re: [dpdk-dev] SIMD Rx/Tx paths
Monday, May 15, 2017 5:26 PM, Thomas Monjalon: > 15/05/2017 16:12, Richardson, Bruce: > > From: Yigit, Ferruh > > > On 5/15/2017 2:15 PM, Bruce Richardson wrote: > > > > On Mon, May 15, 2017 at 02:35:55PM +0200, Thomas Monjalon wrote: > > > >> Hi, > > > >> > > > >> I would like to open a discussion about SIMD code in drivers. > > > >> > > > >> I think we should not have different behaviours or features > > > >> capabilities, in the different code paths of a same driver. > > > >> I suggest to consider such differences as exceptions. > > > >> So we should merge features files (i.e. matrix columns), and > > > >> remove these files: > > > >> > > > >> % ls doc/guides/nics/features/*_vec.ini > > > >> > > > >> doc/guides/nics/features/fm10k_vec.ini > > > >> doc/guides/nics/features/fm10k_vf_vec.ini > > > >> doc/guides/nics/features/i40e_vec.ini > > > >> doc/guides/nics/features/i40e_vf_vec.ini > > > >> doc/guides/nics/features/ixgbe_vec.ini > > > >> doc/guides/nics/features/ixgbe_vf_vec.ini > > > >> doc/guides/nics/features/virtio_vec.ini > > > >> > > > >> If a feature is not supported in all code paths of a driver, it > > > >> must be marked as partially (P) supported. > > > >> > > > >> Then the mid-term goal will be to try removing these inconsistencies. > > > >> > > > >> Opinions/comments? > > > > > > > > Yes, there are inconsistencies, but if they are hidden from the > > > > user, e.g. by having the driver select automatically the most > > > > appropriate path, I don't think we should need to mark the support as > "partial". > > > > Instead, it should be marked as being fully supported, but perhaps > > > > with a note indicating that a performance hit may be experienced > > > > due to the code taking a less-optimised driver path. After all, > > > > especially in the TX code path, a lot of the speed-up comes from > > > > not supporting different features, as well as from the > > > > vectorization. In those cases "closing the gap" may mean losing > > > > performance for those who don't want the features, which is not > > > > acceptable. Any feature support we can add, without affecting > performance, should of course be implemented. > > > > > > I mostly agree with Bruce, except for PMD selection the patch > > > automatically. > > > > > > There is a trade off between feature set and performance, scalar > > > driver favors features and vector driver favors performance, I think > > > good to have both. > > > > > > And I agree that feature support should be added to vector drivers > > > as long as it doesn't effect the performance. > > > > > > Related to the driver auto selecting the path, I concern this may > > > confuse the user, because he may end up a situation he doesn't clear > > > about supported features, I am for more explicit way to select the > > > scalar or vector driver. > > > > > > And for merging the features files, most of the items are already > > > same for scalar and vector drivers, so perhaps we can merge files > > > and use different syntax for features that is different for scalar and > vector: > > > Ys: Yes Scalar [no vector] > > > Yv: Yes Vector [no scalar] > > > Y: Yes Both > > > Ps: Partially Scalar [no vector] > > > Pv: Partially Vector [no scalar] > > > P: Partially Both > > > YsPv, YvPs > > Please remember that there are different vector code paths (SSE/AVX, > NEON, Altivec). > > > For the table, I don't really mind so much what scheme is agreed. For the > user doing the coding, while I can accept that it might be useful to support > explicitly request a vector or scalar driver, I'd definitely prefer the > default > state to remain auto-selection based on features requested. If a user want > TSO, then pick the best driver path that supports TSO, and don't force the > user to read up on what the different paths are! > > I agree. > If we can be sure what the application needs, we can select the best code > path and mark the feature supported. > But can we be sure of the expectations for every features? > How do we know that the application relies on certain packet types (which > are not recognized by some code paths)? This work might help for this [1]. The application will specify on device configuration which Tx and Rx offloads it needs. Knowing that each feature request might affect the performance, the application is expected to choose the most limited set of offloads. The PMD, will then choose accordingly the best data path function which supports all of those offloads, knowing the rest will never be used. [1] http://dpdk.org/ml/archives/dev/2017-May/065077.html
