qede: SR-IOV PF driver support
Subsystems (2)
v4 cover qede: SR-IOV PF driver support
Hi,
This series adds SR-IOV PF pmd driver support to have VF pmd driver work over PF pmd driver instances in order to run the adapter completely under DPDK environment for one of the use cases like ovs-dpdk.
This is very initial bring-up with following testing covered -
* Enable/Disable SR-IOV VFs through igb_uio sysfs hook.
* Load VFs, run fastpath, teardown VFs in hypervisor and guest VM.
* VF FLR flow (in case of VF PCI passthrough to the guest VM)
* Bulletin mechanism tested to communicate link changes to the VFs.
Note that this series is intended for upcoming DPDK release (20.08) Please consider applying this series to dpdk-next-net-mrvl.git
V3->V4:
=======
* Fixed meson/ninja build failure for AMD arch.
* Incorporated cosmetic changes pointed by Gaetan.
V2->V3:
=======
* Instead of adding complete pci_regs.h file from linux, define handful PCI symbols (prefixed with RTE_) in dpdk which are required to implement generic rte_pci_find_next_ext_capability() API (patch #1) and to enable various other qede driver's OSAL_* defines (patch #2) (essentially, removed the patch #1 in v2)
* Fixed meson/ninja compilation issues reported
* Incorporated comments by Gaetan for rte_pci_find_next_ext_capability()
V1->V2: (Incorporated comments from Jerin Jacob)
================================================
* Added rte_pci_regs.h file (copy of linux/pci_regs.h) under lib/librte_pci to remove the dependency of dpdk on user headers
* Added generic API to find PCI extended capability and use that in the drivers, removed individual functions implemented by the drivers
Thanks,
Manish
Manish Chopra (6):
drivers: add generic API to find PCI extended cap net/qede: define PCI config space specific osals net/qede: configure VFs on hardware net/qede: add infrastructure support for VF load net/qede: initialize VF MAC and link net/qede: add VF FLR support
doc/guides/nics/features/qede.ini | 1 + doc/guides/nics/qede.rst | 7 +- drivers/bus/pci/pci_common.c | 43 ++++ drivers/bus/pci/rte_bus_pci.h | 20 ++ drivers/bus/pci/rte_bus_pci_version.map | 6 + drivers/net/ice/ice_ethdev.c | 51 +---- drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c | 48 +---- drivers/net/qede/Makefile | 1 + drivers/net/qede/base/bcm_osal.c | 31 +++ drivers/net/qede/base/bcm_osal.h | 29 ++- drivers/net/qede/base/ecore.h | 7 + drivers/net/qede/base/ecore_dev.c | 6 +- drivers/net/qede/base/ecore_iov_api.h | 3 + drivers/net/qede/base/ecore_sriov.c | 28 +-- drivers/net/qede/base/meson.build | 2 +- drivers/net/qede/meson.build | 1 + drivers/net/qede/qede_ethdev.c | 37 +++- drivers/net/qede/qede_ethdev.h | 1 + drivers/net/qede/qede_if.h | 1 + drivers/net/qede/qede_main.c | 13 +- drivers/net/qede/qede_sriov.c | 219 +++++++++++++++++++++ drivers/net/qede/qede_sriov.h | 22 +++ drivers/raw/ifpga/ifpga_rawdev.c | 6 - lib/librte_pci/rte_pci.h | 35 ++++ 24 files changed, 482 insertions(+), 136 deletions(-) create mode 100644 drivers/net/qede/qede_sriov.c create mode 100644 drivers/net/qede/qede_sriov.h
Signature · 1 line
2.17.1
PATCH v4 1/6 drivers: add generic API to find PCI extended cap
By adding generic API, this patch removes individual functions/defines implemented by drivers to find extended PCI capabilities.
--- drivers/bus/pci/pci_common.c | 43 ++++++++++++++++++ drivers/bus/pci/rte_bus_pci.h | 20 +++++++++ drivers/bus/pci/rte_bus_pci_version.map | 6 +++ drivers/net/ice/ice_ethdev.c | 51 +--------------------- drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c | 48 +------------------- drivers/raw/ifpga/ifpga_rawdev.c | 6 --- lib/librte_pci/rte_pci.h | 16 +++++++ 7 files changed, 89 insertions(+), 101 deletions(-)
drivers/bus/pci/pci_common.c +43 -0
@@ -665,6 +665,49 @@ rte_pci_get_iommu_class(void)return iova_mode; } +off_t +rte_pci_find_next_ext_capability (struct rte_pci_device *dev,uint32_t cap) +{ +off_t offset =RTE_PCI_CFG_SPACE_SIZE ; +uint32_t header; +int ttl; + +/* minimum 8 bytes per capability */ + ttl = (RTE_PCI_CFG_SPACE_EXP_SIZE -RTE_PCI_CFG_SPACE_SIZE ) /8 ; + +if (rte_pci_read_config (dev, &header,4 , offset) <0 ) { +RTE_LOG (ERR ,EAL ,"error in reading extended capabilities\n" ); +return -1 ; + } + +/* +* If we have no capabilities, this is indicated by cap ID, +* cap version and next pointer all being 0. +*/ +if (header ==0 ) +return 0 ; + +while (ttl !=0 ) { +if (RTE_PCI_EXT_CAP_ID (header) == cap) +return offset; + + offset =RTE_PCI_EXT_CAP_NEXT (header); + +if (offset <RTE_PCI_CFG_SPACE_SIZE ) +break ; + +if (rte_pci_read_config (dev, &header,4 , offset) <0 ) { +RTE_LOG (ERR ,EAL , +"error in reading extended capabilities\n" ); +return -1 ; + } + + ttl--; + } + +return 0 ; +} +struct rte_pci_bus rte_pci_bus = { .bus = { .scan = rte_pci_scan,
drivers/bus/pci/rte_bus_pci.h +20 -0
@@ -224,6 +224,26 @@ void rte_pci_unmap_device(struct rte_pci_device *dev);*/ void rte_pci_dump (FILE *f); +/** +* Find device's extended PCI capability. +* +* @param dev +* A pointer to rte_pci_device structure. +* +* @param cap +* Extended capability to be found, which can be any from +* RTE_PCI_EXT_CAP_ID_*, defined in librte_pci. +* +* @return +* > 0: The offset of the next matching extended capability structure +* within the device's PCI configuration space. +* < 0: An error in PCI config space read. +* = 0: Device does not support it. +*/ +__rte_experimental +off_t rte_pci_find_next_ext_capability(struct rte_pci_device *dev, +uint32_t cap); +/** * Register a PCI driver. *
drivers/bus/pci/rte_bus_pci_version.map +6 -0
@@ -16,3 +16,9 @@ DPDK_20.0 { local: *; }; + +EXPERIMENTAL { + global: + + rte_pci_find_next_ext_capability; +};
drivers/net/ice/ice_ethdev.c +2 -49
67 lines not on this page · raw message
drivers/net/nfp/nfpcore/nfp_cpp_pcie_ops.c +2 -46
61 lines not on this page · raw message
drivers/raw/ifpga/ifpga_rawdev.c +0 -6
@@ -41,12 +41,6 @@#include "ifpga_rawdev.h" #include "ipn3ke_rawdev_api.h" -#define RTE_PCI_EXT_CAP_ID_ERR 0x01/* Advanced Error Reporting */ -#define RTE_PCI_CFG_SPACE_SIZE 256 -#define RTE_PCI_CFG_SPACE_EXP_SIZE 4096 -#define RTE_PCI_EXT_CAP_ID (header) (int)(header & 0x0000ffff) -#define RTE_PCI_EXT_CAP_NEXT (header) ((header >> 20) & 0xffc) -#define PCI_VENDOR_ID_INTEL 0x8086/* PCI Device ID */ #define PCIE_DEVICE_ID_PF_INT_5_X 0xBCBD
lib/librte_pci/rte_pci.h +16 -0
@@ -22,6 +22,22 @@ extern "C" {#include <inttypes.h> #include <sys/types.h> + +/* +* Conventional PCI and PCI-X Mode 1 devices have 256 bytes of +* configuration space. PCI-X Mode 2 and PCIe devices have 4096 bytes of +* configuration space. +*/ +#define RTE_PCI_CFG_SPACE_SIZE 256 +#define RTE_PCI_CFG_SPACE_EXP_SIZE 4096 + +/* Extended Capabilities (PCI-X 2.0 and Express) */ +#define RTE_PCI_EXT_CAP_ID (header) (header & 0x0000ffff) +#define RTE_PCI_EXT_CAP_NEXT (header) ((header >> 20) & 0xffc) + +#define RTE_PCI_EXT_CAP_ID_ERR 0x01/* Advanced Error Reporting */ +#define RTE_PCI_EXT_CAP_ID_DSN 0x03/* Device Serial Number */ +/** Formatting string for PCI device identifier: Ex: 0000:00:01.0 */ #define PCI_PRI_FMT "%.4" PRIx32 ":%.2" PRIx8 ":%.2" PRIx8 ".%" PRIx8#define PCI_PRI_STR_SIZE sizeof("XXXXXXXX:XX:XX.X")
Signature · 1 line
2.17.1
Re: [dpdk-dev] [PATCH v4 1/6] drivers: add generic API to find PCI extended cap
Quoted 6 lines · Manish Chopra
By adding generic API, this patch removes individual functions/defines implemented by drivers to find extended PCI capabilities.
Signed-off-by Manish Chopra <[email protected]>Signed-off-by Igor Russkikh <[email protected]>
LGTM,
Signature · 1 line
Gaëtan
Re: [dpdk-dev] [PATCH v4 1/6] drivers: add generic API to find PCI extended cap
Quoted 11 lines · Gaëtan Rivet
Quoted 6 lines · Manish Chopra
By adding generic API, this patch removes individual functions/defines implemented by drivers to find extended PCI capabilities.
Signed-off-by Manish Chopra <[email protected]>Signed-off-by Igor Russkikh <[email protected]>LGTM,
Reviewed-by Gaetan Rivet <[email protected]>
Gaetan Acked this change.
This patch has dependency on other qede patches in the series.
Hi Thomas, David, Ferruh,
Should I take this patch through next-net-mrvl. if not, Could you merge this patch to next-net or main branch so that i can take the remaining patches in this series.
Quoted 3 lines
--
Gaëtan
Re: [dpdk-dev] [PATCH v4 1/6] drivers: add generic API to find PCI extended cap
Quoted 8 lines · Jerin Jacob
Quoted 6 lines · Gaëtan Rivet
Quoted 4 lines · Manish Chopra
By adding generic API, this patch removes individual functions/defines implemented by drivers to find extended PCI capabilities.
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
Signed-off-by Manish Chopra <[email protected]>Signed-off-by Igor Russkikh <[email protected]>LGTM,
Reviewed-by Gaetan Rivet <[email protected]>Gaetan Acked this change.
This patch has dependency on other qede patches in the series.
I can see a v4 and v5 series.
https://patchwork.dpdk.org/project/dpdk/list/?series=11335
https://patchwork.dpdk.org/project/dpdk/list/?series=11421
The v5 has no review tag and the pci patch had a comment from Nithin.
https://patchwork.dpdk.org/patch/75042/#117363
Quoted 6 lines
Hi Thomas, David, Ferruh,
Should I take this patch through next-net-mrvl. if not, Could you merge this patch to next-net or main branch so that i can take the remaining patches in this series.
Let's first sort this out.
Thanks.
Signature · 1 line
David Marchand
Re: [dpdk-dev] [EXT] Re: [PATCH v4 1/6] drivers: add generic API to find PCI extended cap
Quoted 16 lines · Jerin Jacob
Quoted 14 lines · Gaëtan Rivet
Quoted 11 lines · Manish Chopra
Quoted 6 lines
By adding generic API, this patch removes individual functions/defines implemented by drivers to find extended PCI capabilities.
Signed-off-by Manish Chopra <[email protected]>Signed-off-by Igor Russkikh <[email protected]>LGTM,
Reviewed-by Gaetan Rivet <[email protected]>
Quoted 8 lines
Quoted 2 lines
Gaetan Acked this change.
This patch has dependency on other qede patches in the series.I can see a v4 and v5 series.
The v5 has no review tag and the pci patch had a comment from Nithin.
Manish, I think this comment is reasonable, could you please followup with the fix and resubmit v6?
Quoted 6 lines
Quoted 6 lines
Hi Thomas, David, Ferruh,
Should I take this patch through next-net-mrvl. if not, Could you merge this patch to next-net or main branch so that i can take the remaining patches in this series.
Thanks,
Igor
PATCH v4 2/6 net/qede: define PCI config space specific osals
In reply to Manish ChopraThis patch defines various PCI config space access APIs in order to read and find IOV specific PCI capabilities.
With these definitions implemented, it enables the base driver to do SR-IOV specific initialization and HW specific configuration required from PF-PMD driver instance.
--- drivers/net/qede/base/bcm_osal.h | 14 +++++++++----- drivers/net/qede/base/ecore.h | 3 +++ drivers/net/qede/base/ecore_dev.c | 6 +++--- drivers/net/qede/base/ecore_sriov.c | 28 +++++++++++++++------------- drivers/net/qede/base/meson.build | 2 +- drivers/net/qede/qede_main.c | 1 + lib/librte_pci/rte_pci.h | 23 +++++++++++++++++++++-- 7 files changed, 53 insertions(+), 24 deletions(-)
drivers/net/qede/base/bcm_osal.h +9 -5
@@ -21,6 +21,7 @@#include <rte_ether.h> #include <rte_io.h> #include <rte_version.h> +#include <rte_bus_pci.h> /* Forward declaration */ struct ecore_dev ;
@@ -286,11 +287,14 @@ typedef struct osal_list_t {OSAL_LIST_PUSH_HEAD (new_entry, list) /* PCI config space */ - -#define OSAL_PCI_READ_CONFIG_BYTE (dev, address, dst) nothing -#define OSAL_PCI_READ_CONFIG_WORD (dev, address, dst) nothing -#define OSAL_PCI_READ_CONFIG_DWORD (dev, address, dst) nothing -#define OSAL_PCI_FIND_EXT_CAPABILITY (dev, pcie_id) 0 +#define OSAL_PCI_READ_CONFIG_BYTE (dev, address, dst) \ + rte_pci_read_config((dev)->pci_dev, dst, 1, address) +#define OSAL_PCI_READ_CONFIG_WORD (dev, address, dst) \ + rte_pci_read_config((dev)->pci_dev, dst, 2, address) +#define OSAL_PCI_READ_CONFIG_DWORD (dev, address, dst) \ + rte_pci_read_config((dev)->pci_dev, dst, 4, address) +#define OSAL_PCI_FIND_EXT_CAPABILITY (dev, cap) \ + rte_pci_find_next_ext_capability((dev)->pci_dev, cap)#define OSAL_PCI_FIND_CAPABILITY (dev, pcie_id) 0#define OSAL_PCI_WRITE_CONFIG_WORD (dev, address, val) nothing#define OSAL_BAR_SIZE (dev, bar_id) 0
drivers/net/qede/base/ecore.h +3 -0
@@ -937,6 +937,9 @@ struct ecore_dev {struct ecore_dbg_feature dbg_features[DBG_FEATURE_NUM ];struct ecore_dbg_params dbg_params;osal_mutex_t dbg_lock; + +/* DPDK specific ecore field */ +struct rte_pci_device *pci_dev; };enum ecore_hsi_def_type {
drivers/net/qede/base/ecore_dev.c +3 -3
@@ -2787,7 +2787,7 @@ static enum _ecore_status_t ecore_hw_init_chip(struct ecore_dev *p_dev,return ECORE_IO ; } -OSAL_PCI_READ_CONFIG_WORD (p_dev, pos +PCI_EXP_DEVCTL , &ctrl); +OSAL_PCI_READ_CONFIG_WORD (p_dev, pos +RTE_PCI_EXP_DEVCTL , &ctrl); wr_mbs = (ctrl &PCI_EXP_DEVCTL_PAYLOAD ) >>5 ;ecore_wr (p_hwfn, p_ptt,PSWRQ2_REG_WR_MBS0 , wr_mbs);
@@ -5499,9 +5499,9 @@ static enum _ecore_status_t ecore_get_dev_info(struct ecore_hwfn *p_hwfn,u32 tmp;/* Read Vendor Id / Device Id */ -OSAL_PCI_READ_CONFIG_WORD (p_dev,PCICFG_VENDOR_ID_OFFSET , +OSAL_PCI_READ_CONFIG_WORD (p_dev,RTE_PCI_VENDOR_ID , &p_dev->vendor_id); -OSAL_PCI_READ_CONFIG_WORD (p_dev,PCICFG_DEVICE_ID_OFFSET , +OSAL_PCI_READ_CONFIG_WORD (p_dev,RTE_PCI_DEVICE_ID , &p_dev->device_id);/* Determine type */
drivers/net/qede/base/ecore_sriov.c +15 -13
@@ -417,15 +417,16 @@ static enum _ecore_status_t ecore_iov_pci_cfg_info(struct ecore_dev *p_dev)int pos = iov->pos;DP_VERBOSE (p_dev,ECORE_MSG_IOV ,"sriov ext pos %d\n" , pos); -OSAL_PCI_READ_CONFIG_WORD (p_dev, pos +PCI_SRIOV_CTRL , &iov->ctrl); +OSAL_PCI_READ_CONFIG_WORD (p_dev, pos +RTE_PCI_SRIOV_CTRL , &iov->ctrl); +OSAL_PCI_READ_CONFIG_WORD (p_dev, pos +RTE_PCI_SRIOV_TOTAL_VF , + &iov->total_vfs);OSAL_PCI_READ_CONFIG_WORD (p_dev, - pos +PCI_SRIOV_TOTAL_VF , &iov->total_vfs); -OSAL_PCI_READ_CONFIG_WORD (p_dev, - pos +PCI_SRIOV_INITIAL_VF , + pos +RTE_PCI_SRIOV_INITIAL_VF , &iov->initial_vfs); -OSAL_PCI_READ_CONFIG_WORD (p_dev, pos +PCI_SRIOV_NUM_VF , &iov->num_vfs); +OSAL_PCI_READ_CONFIG_WORD (p_dev, pos +RTE_PCI_SRIOV_NUM_VF , + &iov->num_vfs);if (iov->num_vfs) {/* @@@TODO - in future we might want to add an OSAL here to * allow each OS to decide on its own how to act.
@@ -437,20 +438,21 @@ static enum _ecore_status_t ecore_iov_pci_cfg_info(struct ecore_dev *p_dev) }OSAL_PCI_READ_CONFIG_WORD (p_dev, - pos +PCI_SRIOV_VF_OFFSET , &iov->offset); + pos +RTE_PCI_SRIOV_VF_OFFSET , &iov->offset);OSAL_PCI_READ_CONFIG_WORD (p_dev, - pos +PCI_SRIOV_VF_STRIDE , &iov->stride); + pos +RTE_PCI_SRIOV_VF_STRIDE , &iov->stride); -OSAL_PCI_READ_CONFIG_WORD (p_dev, - pos +PCI_SRIOV_VF_DID , &iov->vf_device_id); +OSAL_PCI_READ_CONFIG_WORD (p_dev, pos +RTE_PCI_SRIOV_VF_DID , + &iov->vf_device_id);OSAL_PCI_READ_CONFIG_DWORD (p_dev, - pos +PCI_SRIOV_SUP_PGSIZE , &iov->pgsz); + pos +RTE_PCI_SRIOV_SUP_PGSIZE , &iov->pgsz); -OSAL_PCI_READ_CONFIG_DWORD (p_dev, pos +PCI_SRIOV_CAP , &iov->cap); +OSAL_PCI_READ_CONFIG_DWORD (p_dev, pos +RTE_PCI_SRIOV_CAP , &iov->cap); -OSAL_PCI_READ_CONFIG_BYTE (p_dev, pos +PCI_SRIOV_FUNC_LINK , &iov->link); +OSAL_PCI_READ_CONFIG_BYTE (p_dev, pos +RTE_PCI_SRIOV_FUNC_LINK , + &iov->link);DP_VERBOSE (p_dev,ECORE_MSG_IOV ,"IOV info: nres %d, cap 0x%x," "ctrl 0x%x, total %d, initial %d, num vfs %d, offset %d,"
@@ -669,7 +671,7 @@ enum _ecore_status_t ecore_iov_hw_info(struct ecore_hwfn *p_hwfn)/* Learn the PCI configuration */ pos =OSAL_PCI_FIND_EXT_CAPABILITY (p_hwfn->p_dev, -PCI_EXT_CAP_ID_SRIOV ); +RTE_PCI_EXT_CAP_ID_SRIOV );if (!pos) {DP_VERBOSE (p_hwfn,ECORE_MSG_IOV ,"No PCIe IOV support\n" );return ECORE_SUCCESS ;
drivers/net/qede/base/meson.build +1 -1
@@ -52,6 +52,6 @@ foreach flag: error_cflags endforeach base_lib = static_library('qede_base', sources, - dependencies: static_rte_net, + dependencies: [static_rte_net, static_rte_bus_pci], c_args: c_args) base_objs = base_lib.extract_all_objects()
drivers/net/qede/qede_main.c +1 -0
@@ -37,6 +37,7 @@ static void qed_init_pci(struct ecore_dev *edev, struct rte_pci_device *pci_dev) edev->regview = pci_dev->mem_resource[0 ].addr; edev->doorbells = pci_dev->mem_resource[2 ].addr; edev->db_size = pci_dev->mem_resource[2 ].len; + edev->pci_dev = pci_dev; } static int
lib/librte_pci/rte_pci.h +21 -2
@@ -31,12 +31,31 @@ extern "C" {#define RTE_PCI_CFG_SPACE_SIZE 256#define RTE_PCI_CFG_SPACE_EXP_SIZE 4096 +#define RTE_PCI_VENDOR_ID 0x00/* 16 bits */ +#define RTE_PCI_DEVICE_ID 0x02/* 16 bits */ + +/* PCI Express capability registers */ +#define RTE_PCI_EXP_DEVCTL 8/* Device Control */ +/* Extended Capabilities (PCI-X 2.0 and Express) */ #define RTE_PCI_EXT_CAP_ID (header) (header & 0x0000ffff)#define RTE_PCI_EXT_CAP_NEXT (header) ((header >> 20) & 0xffc) -#define RTE_PCI_EXT_CAP_ID_ERR 0x01/* Advanced Error Reporting */ -#define RTE_PCI_EXT_CAP_ID_DSN 0x03/* Device Serial Number */ +#define RTE_PCI_EXT_CAP_ID_ERR 0x01/* Advanced Error Reporting */ +#define RTE_PCI_EXT_CAP_ID_DSN 0x03/* Device Serial Number */ +#define RTE_PCI_EXT_CAP_ID_SRIOV 0x10/* SR-IOV*/ + +/* Single Root I/O Virtualization */ +#define RTE_PCI_SRIOV_CAP 0x04/* SR-IOV Capabilities */ +#define RTE_PCI_SRIOV_CTRL 0x08/* SR-IOV Control */ +#define RTE_PCI_SRIOV_INITIAL_VF 0x0c/* Initial VFs */ +#define RTE_PCI_SRIOV_TOTAL_VF 0x0e/* Total VFs */ +#define RTE_PCI_SRIOV_NUM_VF 0x10/* Number of VFs */ +#define RTE_PCI_SRIOV_FUNC_LINK 0x12/* Function Dependency Link */ +#define RTE_PCI_SRIOV_VF_OFFSET 0x14/* First VF Offset */ +#define RTE_PCI_SRIOV_VF_STRIDE 0x16/* Following VF Stride */ +#define RTE_PCI_SRIOV_VF_DID 0x1a/* VF Device ID */ +#define RTE_PCI_SRIOV_SUP_PGSIZE 0x1c/* Supported Page Sizes */ /** Formatting string for PCI device identifier: Ex: 0000:00:01.0 */ #define PCI_PRI_FMT "%.4" PRIx32 ":%.2" PRIx8 ":%.2" PRIx8 ".%" PRIx8
Signature · 1 line
2.17.1
PATCH v4 3/6 net/qede: configure VFs on hardware
In reply to Manish ChopraBased on number of VFs enabled at PCI, PF-PMD driver instance enables/configures those VFs from hardware perspective, such that in later patches they could get required HW access to communicate with PFs for slowpath configuration and run the fastpath themsleves.
This patch also add two new qede IOV files [qede_sriov(.c|.h)] under qede directory to add non-base driver IOV APIs/contents there.
--- drivers/net/qede/Makefile | 1 + drivers/net/qede/meson.build | 1 + drivers/net/qede/qede_ethdev.c | 1 + drivers/net/qede/qede_ethdev.h | 1 + drivers/net/qede/qede_if.h | 1 + drivers/net/qede/qede_main.c | 1 + drivers/net/qede/qede_sriov.c | 85 ++++++++++++++++++++++++++++++++++ drivers/net/qede/qede_sriov.h | 9 ++++ 8 files changed, 100 insertions(+) create mode 100644 drivers/net/qede/qede_sriov.c create mode 100644 drivers/net/qede/qede_sriov.h
drivers/net/qede/Makefile +1 -0
@@ -105,5 +105,6 @@ SRCS-$(CONFIG_RTE_LIBRTE_QEDE_PMD) += qede_rxtx.c SRCS-$(CONFIG_RTE_LIBRTE_QEDE_PMD) += qede_filter.c SRCS-$(CONFIG_RTE_LIBRTE_QEDE_PMD) += qede_debug.c SRCS-$(CONFIG_RTE_LIBRTE_QEDE_PMD) += qede_regs.c +SRCS-$(CONFIG_RTE_LIBRTE_QEDE_PMD) += qede_sriov.c include $(RTE_SDK)/mk/rte.lib.mk
drivers/net/qede/meson.build +1 -0
@@ -11,6 +11,7 @@ sources = files( 'qede_rxtx.c', 'qede_debug.c', 'qede_regs.c', + 'qede_sriov.c', ) if cc.has_argument('-Wno-format-nonliteral')
drivers/net/qede/qede_ethdev.c +1 -0
@@ -2700,6 +2700,7 @@ static int qede_common_dev_init(struct rte_eth_dev *eth_dev, bool is_vf) adapter->vxlan.enable = false; adapter->geneve.enable = false; adapter->ipgre.enable = false; + qed_ops->sriov_configure (edev, pci_dev->max_vfs); } DP_INFO(edev, "MAC address : %02x:%02x:%02x:%02x:%02x:%02x\n",
drivers/net/qede/qede_ethdev.h +1 -0
@@ -34,6 +34,7 @@#include "base/ecore_l2.h" #include "base/ecore_vf.h" +#include "qede_sriov.h" #include "qede_logs.h" #include "qede_if.h" #include "qede_rxtx.h"
drivers/net/qede/qede_if.h +1 -0
@@ -82,6 +82,7 @@ struct qed_eth_ops {const struct qed_common_ops *common;int (*fill_dev_info)(struct ecore_dev *edev,struct qed_dev_eth_info *info); +void (*sriov_configure)(struct ecore_dev *edev,int num_vfs); };struct qed_link_params {
drivers/net/qede/qede_main.c +1 -0
@@ -822,6 +822,7 @@ const struct qed_common_ops qed_common_ops_pass = {const struct qed_eth_ops qed_eth_ops_pass = {INIT_STRUCT_FIELD (common, &qed_common_ops_pass),INIT_STRUCT_FIELD (fill_dev_info, &qed_fill_eth_dev_info), +INIT_STRUCT_FIELD (sriov_configure, &qed_sriov_configure), };const struct qed_eth_ops *qed_get_eth_ops (void )
drivers/net/qede/qede_sriov.c new file +85 -0
85 lines not on this page · raw message
drivers/net/qede/qede_sriov.h new file +9 -0
@@ -0,0 +1,9 @@ +/* SPDX-License-Identifier: BSD-3-Clause +* Copyright (c) 2020 Marvell. +* All rights reserved. +* www.marvell.com +*/ + +#include "qede_ethdev.h" + +void qed_sriov_configure (struct ecore_dev *edev,int num_vfs_param);
Signature · 1 line
2.17.1
PATCH v4 4/6 net/qede: add infrastructure support for VF load
In reply to Manish ChopraThis patch adds necessary infrastructure support (required to handle messages from VF and sending ramrod on behalf of VF's configuration request from alarm handler context) to start/load the VF-PMD driver instance on top of PF-PMD driver instance.
--- drivers/net/qede/base/bcm_osal.c | 26 ++++++++++++ drivers/net/qede/base/bcm_osal.h | 11 +++-- drivers/net/qede/base/ecore.h | 4 ++ drivers/net/qede/base/ecore_iov_api.h | 3 ++ drivers/net/qede/qede_ethdev.c | 2 + drivers/net/qede/qede_main.c | 4 +- drivers/net/qede/qede_sriov.c | 61 +++++++++++++++++++++++++++ drivers/net/qede/qede_sriov.h | 16 ++++++- 8 files changed, 121 insertions(+), 6 deletions(-)
drivers/net/qede/base/bcm_osal.c +26 -0
@@ -14,6 +14,32 @@#include "ecore_iov_api.h" #include "ecore_mcp_api.h" #include "ecore_l2_api.h" +#include "../qede_sriov.h" + +int osal_pf_vf_msg (struct ecore_hwfn *p_hwfn) +{ +int rc; + + rc =qed_schedule_iov (p_hwfn,QED_IOV_WQ_MSG_FLAG ); +if (rc) { +DP_VERBOSE (p_hwfn,ECORE_MSG_IOV , +"Failed to schedule alarm handler rc=%d\n" , rc); + } + +return rc; +} + +void osal_poll_mode_dpc (osal_int_ptr_t hwfn_cookie) +{ +struct ecore_hwfn *p_hwfn = (struct ecore_hwfn *)hwfn_cookie; + +if (!p_hwfn) +return ; + +OSAL_SPIN_LOCK (&p_hwfn->spq_lock); +ecore_int_sp_dpc ((osal_int_ptr_t)(p_hwfn)); +OSAL_SPIN_UNLOCK (&p_hwfn->spq_lock); +}/* Array of memzone pointers */ static const struct rte_memzone *ecore_mz_mapping[RTE_MAX_MEMZONE ];
drivers/net/qede/base/bcm_osal.h +8 -3
@@ -178,9 +178,12 @@ typedef pthread_mutex_t osal_mutex_t;/* DPC */ +void osal_poll_mode_dpc (osal_int_ptr_t hwfn_cookie);#define OSAL_DPC_ALLOC (hwfn) OSAL_ALLOC(hwfn, GFP, sizeof(osal_dpc_t)) -#define OSAL_DPC_INIT (dpc, hwfn) nothing -#define OSAL_POLL_MODE_DPC (hwfn) nothing +#define OSAL_DPC_INIT (dpc, hwfn) \ + OSAL_SPIN_LOCK_INIT(&(hwfn)->spq_lock) +#define OSAL_POLL_MODE_DPC (hwfn) \ + osal_poll_mode_dpc((osal_int_ptr_t)(p_hwfn))#define OSAL_DPC_SYNC (hwfn) nothing/* Lists */
@@ -345,10 +348,12 @@ u32 qede_find_first_zero_bit(u32 *bitmap, u32 length);/* SR-IOV channel */ +int osal_pf_vf_msg (struct ecore_hwfn *p_hwfn);#define OSAL_VF_FLR_UPDATE (hwfn) nothing#define OSAL_VF_SEND_MSG2PF (dev, done, msg, reply_addr, msg_size, reply_size) 0#define OSAL_VF_CQE_COMPLETION (_dev_p, _cqe, _protocol) (0) -#define OSAL_PF_VF_MSG (hwfn, vfid) 0 +#define OSAL_PF_VF_MSG (hwfn, vfid) \ + osal_pf_vf_msg(hwfn)#define OSAL_PF_VF_MALICIOUS (hwfn, vfid) nothing#define OSAL_IOV_CHK_UCAST (hwfn, vfid, params) 0#define OSAL_IOV_POST_START_VPORT (hwfn, vf, vport_id, opaque_fid) nothing
drivers/net/qede/base/ecore.h +4 -0
@@ -714,6 +714,10 @@ struct ecore_hwfn {/* @DPDK */ struct ecore_ptt *p_arfs_ptt; + +/* DPDK specific, not the part of vanilla ecore */ +osal_spinlock_t spq_lock; +u32 iov_task_flags; };enum ecore_mf_mode {
drivers/net/qede/base/ecore_iov_api.h +3 -0
@@ -14,6 +14,9 @@#define ECORE_ETH_VF_NUM_VLAN_FILTERS 2#define ECORE_VF_ARRAY_LENGTH (3) +#define ECORE_VF_ARRAY_GET_VFID (arr, vfid) \ + (((arr)[(vfid) / 64]) & (1ULL << ((vfid) % 64))) +#define IS_VF (p_dev) ((p_dev)->b_is_vf)#define IS_PF (p_dev) (!((p_dev)->b_is_vf))#ifdef CONFIG_ECORE_SRIOV
drivers/net/qede/qede_ethdev.c +2 -0
@@ -281,7 +281,9 @@ qede_fw_version_get(struct rte_eth_dev *dev, char *fw_ver, size_t fw_size)static void qede_interrupt_action (struct ecore_hwfn *p_hwfn) { +OSAL_SPIN_LOCK (&p_hwfn->spq_lock);ecore_int_sp_dpc ((osal_int_ptr_t)(p_hwfn)); +OSAL_SPIN_UNLOCK (&p_hwfn->spq_lock); } static void
drivers/net/qede/qede_main.c +3 -1
@@ -221,7 +221,9 @@ static void qed_stop_iov_task(struct ecore_dev *edev)for_each_hwfn (edev, i) { p_hwfn = &edev->hwfns[i]; -if (!IS_PF (edev)) +if (IS_PF (edev)) +rte_eal_alarm_cancel (qed_iov_pf_task, p_hwfn); +else rte_eal_alarm_cancel (qede_vf_task, p_hwfn); } }
drivers/net/qede/qede_sriov.c +61 -0
70 lines not on this page · raw message
drivers/net/qede/qede_sriov.h +14 -2
@@ -4,6 +4,18 @@* www.marvell.com */ -#include "qede_ethdev.h" -void qed_sriov_configure (struct ecore_dev *edev,int num_vfs_param); + +enum qed_iov_wq_flag { +QED_IOV_WQ_MSG_FLAG , +QED_IOV_WQ_SET_UNICAST_FILTER_FLAG , +QED_IOV_WQ_BULLETIN_UPDATE_FLAG , +QED_IOV_WQ_STOP_WQ_FLAG , +QED_IOV_WQ_FLR_FLAG , +QED_IOV_WQ_TRUST_FLAG , +QED_IOV_WQ_VF_FORCE_LINK_QUERY_FLAG , +QED_IOV_WQ_DB_REC_HANDLER , +}; + +int qed_schedule_iov (struct ecore_hwfn *p_hwfn,enum qed_iov_wq_flag flag); +void qed_iov_pf_task (void *arg);
Signature · 1 line
2.17.1
PATCH v4 5/6 net/qede: initialize VF MAC and link
In reply to Manish ChopraThis patch configures VFs with random mac if no MAC is provided by the PF/bulletin. This also adds required bulletin APIs by PF-PMD driver to communicate LINK properties/changes to the VFs through bulletin update mechanism.
With these changes, VF-PMD instance is able to run fastpath over PF-PMD driver instance.
--- drivers/net/qede/qede_ethdev.c | 34 ++++++++++++++++++++- drivers/net/qede/qede_main.c | 7 ++++- drivers/net/qede/qede_sriov.c | 55 ++++++++++++++++++++++++++++++++++ drivers/net/qede/qede_sriov.h | 1 + 4 files changed, 95 insertions(+), 2 deletions(-)
drivers/net/qede/qede_ethdev.c +33 -1
@@ -2479,6 +2479,24 @@ static void qede_update_pf_params(struct ecore_dev *edev) qed_ops->common->update_pf_params (edev, &pf_params); } +static void qede_generate_random_mac_addr (struct rte_ether_addr *mac_addr) +{ +uint64_t random; + +/* Set Organizationally Unique Identifier (OUI) prefix. */ + mac_addr->addr_bytes[0 ] =0x00 ; + mac_addr->addr_bytes[1 ] =0x09 ; + mac_addr->addr_bytes[2 ] =0xC0 ; + +/* Force indication of locally assigned MAC address. */ + mac_addr->addr_bytes[0 ] |=RTE_ETHER_LOCAL_ADMIN_ADDR ; + +/* Generate the last 3 bytes of the MAC address with a random number. */ + random =rte_rand (); + +memcpy (&mac_addr->addr_bytes[3 ], &random,3 ); +} +static int qede_common_dev_init (struct rte_eth_dev *eth_dev,bool is_vf) {struct rte_pci_device *pci_dev;
@@ -2491,7 +2509,7 @@ static int qede_common_dev_init(struct rte_eth_dev *eth_dev, bool is_vf)uint8_t bulletin_change;uint8_t vf_mac[RTE_ETHER_ADDR_LEN ];uint8_t is_mac_forced; -bool is_mac_exist; +bool is_mac_exist = false;/* Fix up ecore debug level */ uint32_t dp_module = ~0 & ~ECORE_MSG_HW ;uint8_t dp_level =ECORE_LEVEL_VERBOSE ;
@@ -2669,6 +2687,20 @@ static int qede_common_dev_init(struct rte_eth_dev *eth_dev, bool is_vf)DP_ERR (edev,"No VF macaddr assigned\n" ); } } + +/* If MAC doesn't exist from PF, generate random one */ +if (!is_mac_exist) { +struct rte_ether_addr *mac_addr; + + mac_addr = (struct rte_ether_addr *)&vf_mac; +qede_generate_random_mac_addr (mac_addr); + +rte_ether_addr_copy (mac_addr, + ð_dev->data->mac_addrs[0 ]); + +rte_ether_addr_copy (ð_dev->data->mac_addrs[0 ], + &adapter->primary_mac); + } } eth_dev->dev_ops = (is_vf) ? &qede_eth_vf_dev_ops : &qede_eth_dev_ops;
drivers/net/qede/qede_main.c +6 -1
@@ -651,10 +651,15 @@ void qed_link_update(struct ecore_hwfn *hwfn)struct ecore_dev *edev = hwfn->p_dev;struct qede_dev *qdev = (struct qede_dev *)edev;struct rte_eth_dev *dev = (struct rte_eth_dev *)qdev->ethdev; +int rc; + + rc =qede_link_update (dev,0 ); +qed_inform_vf_link_state (hwfn); -if (!qede_link_update (dev,0 )) +if (!rc) {_rte_eth_dev_callback_process (dev,RTE_ETH_EVENT_INTR_LSC ,NULL ); + } }static int qed_drain (struct ecore_dev *edev)
drivers/net/qede/qede_sriov.c +55 -0
70 lines not on this page · raw message
drivers/net/qede/qede_sriov.h +1 -0
@@ -17,5 +17,6 @@ enum qed_iov_wq_flag { QED_IOV_WQ_DB_REC_HANDLER, }; +void qed_inform_vf_link_state (struct ecore_hwfn *hwfn);int qed_schedule_iov (struct ecore_hwfn *p_hwfn,enum qed_iov_wq_flag flag);void qed_iov_pf_task (void *arg);
Signature · 1 line
2.17.1
PATCH v4 6/6 net/qede: add VF FLR support
In reply to Manish ChopraThis patch adds required bit to handle VF FLR indication from Management FW (MFW) of the device
With that VFs were able to load in VM (VF attached as PCI passthrough to the guest VM) followed by FLR successfully
Updated the docs/guides with the feature support
--- doc/guides/nics/features/qede.ini | 1 + doc/guides/nics/qede.rst | 7 +------ drivers/net/qede/base/bcm_osal.c | 5 +++++ drivers/net/qede/base/bcm_osal.h | 4 +++- drivers/net/qede/qede_sriov.c | 18 ++++++++++++++++++ 5 files changed, 28 insertions(+), 7 deletions(-)
doc/guides/nics/features/qede.ini +1 -0
@@ -32,6 +32,7 @@ Basic stats = Y Extended stats = Y Stats per queue = Y Registers dump = Y +SR-IOV = Y Multiprocess aware = Y Linux UIO = Y Linux VFIO = Y
doc/guides/nics/qede.rst +1 -6
@@ -34,18 +34,13 @@ Supported Features - VLAN offload - Filtering and stripping - N-tuple filter and flow director (limited support) - NPAR (NIC Partitioning) -- SR-IOV VF +- SR-IOV PF and VF - GRE Tunneling offload - GENEVE Tunneling offload - VXLAN Tunneling offload - MPLSoUDP Tx Tunneling offload - Generic flow API -Non-supported Features ----------------------- - -- SR-IOV PF - Co-existence considerations ---------------------------
drivers/net/qede/base/bcm_osal.c +5 -0
@@ -29,6 +29,11 @@ int osal_pf_vf_msg(struct ecore_hwfn *p_hwfn)return rc; } +void osal_vf_flr_update (struct ecore_hwfn *p_hwfn) +{ +qed_schedule_iov (p_hwfn,QED_IOV_WQ_FLR_FLAG ); +} +void osal_poll_mode_dpc (osal_int_ptr_t hwfn_cookie) {struct ecore_hwfn *p_hwfn = (struct ecore_hwfn *)hwfn_cookie;
drivers/net/qede/base/bcm_osal.h +3 -1
@@ -349,7 +349,9 @@ u32 qede_find_first_zero_bit(u32 *bitmap, u32 length);/* SR-IOV channel */ int osal_pf_vf_msg (struct ecore_hwfn *p_hwfn); -#define OSAL_VF_FLR_UPDATE (hwfn) nothing +void osal_vf_flr_update (struct ecore_hwfn *p_hwfn); +#define OSAL_VF_FLR_UPDATE (hwfn) \ + osal_vf_flr_update(hwfn)#define OSAL_VF_SEND_MSG2PF (dev, done, msg, reply_addr, msg_size, reply_size) 0#define OSAL_VF_CQE_COMPLETION (_dev_p, _cqe, _protocol) (0)#define OSAL_PF_VF_MSG (hwfn, vfid) \
drivers/net/qede/qede_sriov.c +18 -0
@@ -151,6 +151,7 @@ static void qed_handle_bulletin_post(struct ecore_hwfn *hwfn)void qed_iov_pf_task (void *arg) {struct ecore_hwfn *p_hwfn = arg; +int rc;if (OSAL_GET_BIT (QED_IOV_WQ_MSG_FLAG , &p_hwfn->iov_task_flags)) {OSAL_CLEAR_BIT (QED_IOV_WQ_MSG_FLAG , &p_hwfn->iov_task_flags);
@@ -163,6 +164,23 @@ void qed_iov_pf_task(void *arg) &p_hwfn->iov_task_flags);qed_handle_bulletin_post (p_hwfn); } + +if (OSAL_GET_BIT (QED_IOV_WQ_FLR_FLAG , &p_hwfn->iov_task_flags)) { +struct ecore_ptt *p_ptt =ecore_ptt_acquire (p_hwfn); + +OSAL_CLEAR_BIT (QED_IOV_WQ_FLR_FLAG , &p_hwfn->iov_task_flags); + +if (!p_ptt) { +qed_schedule_iov (p_hwfn,QED_IOV_WQ_FLR_FLAG ); +return ; + } + + rc =ecore_iov_vf_flr_cleanup (p_hwfn, p_ptt); +if (rc) +qed_schedule_iov (p_hwfn,QED_IOV_WQ_FLR_FLAG ); + +ecore_ptt_release (p_hwfn, p_ptt); + } }int qed_schedule_iov (struct ecore_hwfn *p_hwfn,enum qed_iov_wq_flag flag)
Signature · 1 line
2.17.1