net: mana: Add support for EQ sharing
Haiyang Zhang
maintainerv2 cover net: mana: Add support for EQ sharing
The existing code uses (1 + #vPorts * #Queues) MSIXs, which may exceed the device limit.
Support EQ sharing, so that multiple vPorts can share the same set of MSIXs.
Haiyang Zhang (3):
net: mana: Move NAPI from EQ to CQ
net: mana: Add support for EQ sharing
net: mana: Add WARN_ON_ONCE in case of CQE read overflow
drivers/net/ethernet/microsoft/mana/gdma.h | 32 ++-- .../net/ethernet/microsoft/mana/gdma_main.c | 88 +++------- .../net/ethernet/microsoft/mana/hw_channel.c | 2 +- drivers/net/ethernet/microsoft/mana/mana.h | 29 ++-- drivers/net/ethernet/microsoft/mana/mana_en.c | 162 ++++++++++-------- 5 files changed, 153 insertions(+), 160 deletions(-)
Signature · 1 line
2.25.1
Haiyang Zhang
maintainerPATCH v2 1/3 net: mana: Move NAPI from EQ to CQ
The existing code has NAPI threads polling on EQ directly. To prepare for EQ sharing among vPorts, move NAPI from EQ to CQ so that one EQ can serve multiple CQs from different vPorts.
The "arm bit" is only set when CQ processing is completed to reduce the number of EQ entries, which in turn reduce the number of interrupts on EQ.
---
v2: Updates suggested by Dexuan Cui <[email protected]>
--- drivers/net/ethernet/microsoft/mana/gdma.h | 9 +- .../net/ethernet/microsoft/mana/gdma_main.c | 55 +--------- .../net/ethernet/microsoft/mana/hw_channel.c | 2 +- drivers/net/ethernet/microsoft/mana/mana.h | 11 +- drivers/net/ethernet/microsoft/mana/mana_en.c | 100 +++++++++++------- 5 files changed, 74 insertions(+), 103 deletions(-)
drivers/net/ethernet/microsoft/mana/gdma.h +1 -8
@@ -239,10 +239,8 @@ struct gdma_event {struct gdma_queue ; -#define CQE_POLLING_BUFFER 512struct mana_eq {struct gdma_queue *eq; -struct gdma_comp cqe_poll[CQE_POLLING_BUFFER ]; }; typedef void gdma_eq_callback(void *context, struct gdma_queue *q,
@@ -291,11 +289,6 @@ struct gdma_queue {unsigned int msix_index;u32 log2_throttle_limit; - -/* NAPI data */ -struct napi_struct napi; -int work_done; -int budget; } eq;struct {
@@ -406,7 +399,7 @@ void mana_gd_destroy_queue(struct gdma_context *gc, struct gdma_queue *queue);int mana_gd_poll_cq (struct gdma_queue *cq,struct gdma_comp *comp,int num_cqe); -void mana_gd_arm_cq (struct gdma_queue *cq); +void mana_gd_ring_cq (struct gdma_queue *cq,u8 arm_bit);struct gdma_wqe {u32 reserved :24 ;
drivers/net/ethernet/microsoft/mana/gdma_main.c +4 -51
104 lines not on this page · raw message
drivers/net/ethernet/microsoft/mana/hw_channel.c +1 -1
@@ -304,7 +304,7 @@ static void mana_hwc_comp_event(void *ctx, struct gdma_queue *q_self) &comp_data); } -mana_gd_arm_cq (q_self); +mana_gd_ring_cq (q_self,SET_ARM_BIT ); }static void mana_hwc_destroy_cq (struct gdma_context *gc,struct hwc_cq *hwc_cq)
drivers/net/ethernet/microsoft/mana/mana.h +9 -2
@@ -225,6 +225,8 @@ struct mana_tx_comp_oob {struct mana_rxq ; +#define CQE_POLLING_BUFFER 512 +struct mana_cq {struct gdma_queue *gdma_cq;
@@ -244,8 +246,13 @@ struct mana_cq {*/ struct mana_txq *txq; -/* Pointer to a buffer which the CQ handler can copy the CQE's into. */ -struct gdma_comp *gdma_comp_buf; +/* Buffer which the CQ handler can copy the CQE's into. */ +struct gdma_comp gdma_comp_buf[CQE_POLLING_BUFFER ]; + +/* NAPI data */ +struct napi_struct napi; +int work_done; +int budget; };#define GDMA_MAX_RQE_SGES 15
drivers/net/ethernet/microsoft/mana/mana_en.c +59 -41
225 lines not on this page · raw message
Signature · 1 line
2.25.1
Haiyang Zhang
maintainerPATCH v2 2/3 net: mana: Add support for EQ sharing
In reply to Haiyang ZhangThe existing code uses (1 + #vPorts * #Queues) MSIXs, which may exceed the device limit.
Support EQ sharing, so that multiple vPorts (NICs) can share the same set of MSIXs.
And, report the EQ-sharing capability bit to the host, which means the host can potentially offer more vPorts and queues to the VM.
Also update the resource limit checking and error handling for better robustness.
Now, we support up to 256 virtual ports per VF (it was 16/VF), and support up to 64 queues per vPort (it was 16).
---
v2: Updates suggested by Dexuan Cui <[email protected]>
Limit max_num_queues to (gc->num_msix_usable - 1)
Add BUILD_BUG_ON(MAX_PORTS_IN_MANA_DEV * 2 * GDMA_EQE_SIZE > EQ_SIZE) More detailed descriptions
--- drivers/net/ethernet/microsoft/mana/gdma.h | 23 ++++--- .../net/ethernet/microsoft/mana/gdma_main.c | 31 +++++++--- drivers/net/ethernet/microsoft/mana/mana.h | 18 +++--- drivers/net/ethernet/microsoft/mana/mana_en.c | 62 +++++++++---------- 4 files changed, 78 insertions(+), 56 deletions(-)
drivers/net/ethernet/microsoft/mana/gdma.h +16 -7
@@ -312,9 +312,6 @@ struct gdma_queue_spec {void *context;unsigned long log2_throttle_limit; - -/* Only used by the MANA device. */ -struct net_device *ndev; } eq;struct {
@@ -489,16 +486,28 @@ enum { GDMA_PROTOCOL_LAST = GDMA_PROTOCOL_V1, }; +#define GDMA_DRV_CAP_FLAG_1_EQ_SHARING_MULTI_VPORT BIT(0) + +#define GDMA_DRV_CAP_FLAGS1 GDMA_DRV_CAP_FLAG_1_EQ_SHARING_MULTI_VPORT + +#define GDMA_DRV_CAP_FLAGS2 0 + +#define GDMA_DRV_CAP_FLAGS3 0 + +#define GDMA_DRV_CAP_FLAGS4 0 +struct gdma_verify_ver_req {struct gdma_req_hdr hdr;/* Mandatory fields required for protocol establishment */ u64 protocol_ver_min;u64 protocol_ver_max; -u64 drv_cap_flags1; -u64 drv_cap_flags2; -u64 drv_cap_flags3; -u64 drv_cap_flags4; + +/* Gdma Driver Capability Flags */ +u64 gd_drv_cap_flags1; +u64 gd_drv_cap_flags2; +u64 gd_drv_cap_flags3; +u64 gd_drv_cap_flags4;/* Advisory fields */ u64 drv_ver;
drivers/net/ethernet/microsoft/mana/gdma_main.c +22 -9
@@ -67,6 +67,10 @@ static int mana_gd_query_max_resources(struct pci_dev *pdev)if (gc->max_num_queues > resp.max_rq) gc->max_num_queues = resp.max_rq; +/* The Hardware Channel (HWC) used 1 MSI-X */ +if (gc->max_num_queues > gc->num_msix_usable -1 ) + gc->max_num_queues = gc->num_msix_usable -1 ; +return 0 ; }
@@ -384,28 +388,31 @@ static int mana_gd_register_irq(struct gdma_queue *queue,struct gdma_resource *r;unsigned int msi_index;unsigned long flags; -int err; +struct device *dev; +int err =0 ; gc = gd->gdma_context; r = &gc->msix_resource; + dev = gc->dev;spin_lock_irqsave (&r->lock, flags); msi_index =find_first_zero_bit (r->map, r->size); -if (msi_index >= r->size) { +if (msi_index >= r->size || msi_index >= gc->num_msix_usable) { err = -ENOSPC ; }else {bitmap_set (r->map, msi_index,1 ); queue->eq.msix_index = msi_index; - err =0 ; }spin_unlock_irqrestore (&r->lock, flags); -if (err) -return err; +if (err) { +dev_err (dev,"Register IRQ err:%d, msi:%u rsize:%u, nMSI:%u" , + err, msi_index, r->size, gc->num_msix_usable); -WARN_ON (msi_index >= gc->num_msix_usable); +return err; + } gic = &gc->irq_contexts[msi_index];
@@ -836,6 +843,11 @@ int mana_gd_verify_vf_version(struct pci_dev *pdev) req.protocol_ver_min =GDMA_PROTOCOL_FIRST ; req.protocol_ver_max =GDMA_PROTOCOL_LAST ; + req.gd_drv_cap_flags1 =GDMA_DRV_CAP_FLAGS1 ; + req.gd_drv_cap_flags2 =GDMA_DRV_CAP_FLAGS2 ; + req.gd_drv_cap_flags3 =GDMA_DRV_CAP_FLAGS3 ; + req.gd_drv_cap_flags4 =GDMA_DRV_CAP_FLAGS4 ; + err =mana_gd_send_request (gc,sizeof (req), &req,sizeof (resp), &resp);if (err || resp.hdr.status) { dev_err(gc->dev, "VfVerifyVersionOutput: %d, status=0x%x\n",
@@ -1154,10 +1166,8 @@ static int mana_gd_setup_irqs(struct pci_dev *pdev)if (max_queues_per_port >MANA_MAX_NUM_QUEUES ) max_queues_per_port =MANA_MAX_NUM_QUEUES ; - max_irqs = max_queues_per_port *MAX_PORTS_IN_MANA_DEV ; -/* Need 1 interrupt for the Hardware communication Channel (HWC) */ - max_irqs++; + max_irqs = max_queues_per_port +1 ; nvec =pci_alloc_irq_vectors (pdev,2 , max_irqs,PCI_IRQ_MSIX );if (nvec <0 )
@@ -1244,6 +1254,9 @@ static int mana_gd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)int bar =0 ;int err; +/* Each port has 2 CQs, each CQ has at most 1 EQE at a time */ +BUILD_BUG_ON (2 *MAX_PORTS_IN_MANA_DEV *GDMA_EQE_SIZE >EQ_SIZE ); + err =pci_enable_device (pdev);if (err)return -ENXIO ;
drivers/net/ethernet/microsoft/mana/mana.h +9 -9
@@ -46,7 +46,7 @@ enum TRI_STATE {#define EQ_SIZE (8 * PAGE_SIZE)#define LOG2_EQ_THROTTLE 3 -#define MAX_PORTS_IN_MANA_DEV 16 +#define MAX_PORTS_IN_MANA_DEV 256struct mana_stats {u64 packets;
@@ -322,6 +322,8 @@ struct mana_context {u16 num_ports; +struct mana_eq *eqs; +struct net_device *ports[MAX_PORTS_IN_MANA_DEV ]; };
@@ -331,8 +333,6 @@ struct mana_port_context {u8 mac_addr[ETH_ALEN ]; -struct mana_eq *eqs; -enum TRI_STATE rss_state;mana_handle_t default_rxobj;
@@ -402,11 +402,11 @@ enum mana_command_code {struct mana_query_device_cfg_req {struct gdma_req_hdr hdr; -/* Driver Capability flags */ -u64 drv_cap_flags1; -u64 drv_cap_flags2; -u64 drv_cap_flags3; -u64 drv_cap_flags4; +/* MANA Nic Driver Capability flags */ +u64 mn_drv_cap_flags1; +u64 mn_drv_cap_flags2; +u64 mn_drv_cap_flags3; +u64 mn_drv_cap_flags4;u32 proto_major_ver;u32 proto_minor_ver;
@@ -523,7 +523,7 @@ struct mana_cfg_rx_steer_resp {struct gdma_resp_hdr hdr; };/* HW DATA */ -#define MANA_MAX_NUM_QUEUES 16 +#define MANA_MAX_NUM_QUEUES 64#define MANA_SHORT_VPORT_OFFSET_MAX ((1U << 8) - 1)
drivers/net/ethernet/microsoft/mana/mana_en.c +31 -31
164 lines not on this page · raw message
Signature · 1 line
2.25.1
Haiyang Zhang
maintainerPATCH v2 3/3 net: mana: Add WARN_ON_ONCE in case of CQE read overflow
In reply to Haiyang ZhangThis is not an expected case normally.
Add WARN_ON_ONCE in case of CQE read overflow, instead of failing silently.
---
v2: Add this warning suggested by Dexuan Cui <[email protected]>
--- drivers/net/ethernet/microsoft/mana/gdma_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/microsoft/mana/gdma_main.c +1 -1
@@ -1093,7 +1093,7 @@ static int mana_gd_read_cqe(struct gdma_queue *cq, struct gdma_comp *comp) new_bits = (cq->head / num_cqe) &GDMA_CQE_OWNER_MASK ;/* Return -1 if overflow detected. */ -if (owner_bits != new_bits) +if (WARN_ON_ONCE (owner_bits != new_bits))return -1 ; comp->wq_num = cqe->cqe_info.wq_num;
Signature · 1 line
2.25.1
This series was applied to netdev/net-next.git (refs/heads/master):
Hello:
This series was applied to netdev/net-next.git (refs/heads/master):
Quoted 6 lines · On Tue, 24 Aug 2021 09:45:58 -0700 you wrote:
The existing code uses (1 + #vPorts * #Queues) MSIXs, which may exceed the device limit.
Support EQ sharing, so that multiple vPorts can share the same set of MSIXs.
Quoted 6 lines
Haiyang Zhang (3):
net: mana: Move NAPI from EQ to CQ
net: mana: Add support for EQ sharing
net: mana: AddWARN_ON_ONCEin case of CQE read overflow[...]
Here is the summary with links:
- [V2,net-next,1/3] net: mana: Move NAPI from EQ to CQ
https://git.kernel.org/netdev/net-next/c/e1b5683ff62e - [V2,net-next,2/3] net: mana: Add support for EQ sharing https://git.kernel.org/netdev/net-next/c/1e2d0824a9c3 - [V2,net-next,3/3] net: mana: Add WARN_ON_ONCE in case of CQE read overflow https://git.kernel.org/netdev/net-next/c/c1a3e9f98dde
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html