Commit e93826d3 authored by David S. Miller's avatar David S. Miller

Merge branch 'mana-EQ-sharing'

Haiyang Zhang says:

====================
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.
====================
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parents 807d1032 c1a3e9f9
...@@ -239,10 +239,8 @@ struct gdma_event { ...@@ -239,10 +239,8 @@ struct gdma_event {
struct gdma_queue; struct gdma_queue;
#define CQE_POLLING_BUFFER 512
struct mana_eq { struct mana_eq {
struct gdma_queue *eq; struct gdma_queue *eq;
struct gdma_comp cqe_poll[CQE_POLLING_BUFFER];
}; };
typedef void gdma_eq_callback(void *context, struct gdma_queue *q, typedef void gdma_eq_callback(void *context, struct gdma_queue *q,
...@@ -291,11 +289,6 @@ struct gdma_queue { ...@@ -291,11 +289,6 @@ struct gdma_queue {
unsigned int msix_index; unsigned int msix_index;
u32 log2_throttle_limit; u32 log2_throttle_limit;
/* NAPI data */
struct napi_struct napi;
int work_done;
int budget;
} eq; } eq;
struct { struct {
...@@ -319,9 +312,6 @@ struct gdma_queue_spec { ...@@ -319,9 +312,6 @@ struct gdma_queue_spec {
void *context; void *context;
unsigned long log2_throttle_limit; unsigned long log2_throttle_limit;
/* Only used by the MANA device. */
struct net_device *ndev;
} eq; } eq;
struct { struct {
...@@ -406,7 +396,7 @@ void mana_gd_destroy_queue(struct gdma_context *gc, struct gdma_queue *queue); ...@@ -406,7 +396,7 @@ 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); 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 { struct gdma_wqe {
u32 reserved :24; u32 reserved :24;
...@@ -496,16 +486,28 @@ enum { ...@@ -496,16 +486,28 @@ enum {
GDMA_PROTOCOL_LAST = GDMA_PROTOCOL_V1, 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_verify_ver_req {
struct gdma_req_hdr hdr; struct gdma_req_hdr hdr;
/* Mandatory fields required for protocol establishment */ /* Mandatory fields required for protocol establishment */
u64 protocol_ver_min; u64 protocol_ver_min;
u64 protocol_ver_max; u64 protocol_ver_max;
u64 drv_cap_flags1;
u64 drv_cap_flags2; /* Gdma Driver Capability Flags */
u64 drv_cap_flags3; u64 gd_drv_cap_flags1;
u64 drv_cap_flags4; u64 gd_drv_cap_flags2;
u64 gd_drv_cap_flags3;
u64 gd_drv_cap_flags4;
/* Advisory fields */ /* Advisory fields */
u64 drv_ver; u64 drv_ver;
......
...@@ -67,6 +67,10 @@ static int mana_gd_query_max_resources(struct pci_dev *pdev) ...@@ -67,6 +67,10 @@ static int mana_gd_query_max_resources(struct pci_dev *pdev)
if (gc->max_num_queues > resp.max_rq) if (gc->max_num_queues > resp.max_rq)
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; return 0;
} }
...@@ -267,7 +271,7 @@ void mana_gd_wq_ring_doorbell(struct gdma_context *gc, struct gdma_queue *queue) ...@@ -267,7 +271,7 @@ void mana_gd_wq_ring_doorbell(struct gdma_context *gc, struct gdma_queue *queue)
queue->id, queue->head * GDMA_WQE_BU_SIZE, 1); queue->id, queue->head * GDMA_WQE_BU_SIZE, 1);
} }
void mana_gd_arm_cq(struct gdma_queue *cq) void mana_gd_ring_cq(struct gdma_queue *cq, u8 arm_bit)
{ {
struct gdma_context *gc = cq->gdma_dev->gdma_context; struct gdma_context *gc = cq->gdma_dev->gdma_context;
...@@ -276,7 +280,7 @@ void mana_gd_arm_cq(struct gdma_queue *cq) ...@@ -276,7 +280,7 @@ void mana_gd_arm_cq(struct gdma_queue *cq)
u32 head = cq->head % (num_cqe << GDMA_CQE_OWNER_BITS); u32 head = cq->head % (num_cqe << GDMA_CQE_OWNER_BITS);
mana_gd_ring_doorbell(gc, cq->gdma_dev->doorbell, cq->type, cq->id, mana_gd_ring_doorbell(gc, cq->gdma_dev->doorbell, cq->type, cq->id,
head, SET_ARM_BIT); head, arm_bit);
} }
static void mana_gd_process_eqe(struct gdma_queue *eq) static void mana_gd_process_eqe(struct gdma_queue *eq)
...@@ -339,7 +343,6 @@ static void mana_gd_process_eq_events(void *arg) ...@@ -339,7 +343,6 @@ static void mana_gd_process_eq_events(void *arg)
struct gdma_queue *eq = arg; struct gdma_queue *eq = arg;
struct gdma_context *gc; struct gdma_context *gc;
struct gdma_eqe *eqe; struct gdma_eqe *eqe;
unsigned int arm_bit;
u32 head, num_eqe; u32 head, num_eqe;
int i; int i;
...@@ -370,91 +373,53 @@ static void mana_gd_process_eq_events(void *arg) ...@@ -370,91 +373,53 @@ static void mana_gd_process_eq_events(void *arg)
eq->head++; eq->head++;
} }
/* Always rearm the EQ for HWC. For MANA, rearm it when NAPI is done. */
if (mana_gd_is_hwc(eq->gdma_dev)) {
arm_bit = SET_ARM_BIT;
} else if (eq->eq.work_done < eq->eq.budget &&
napi_complete_done(&eq->eq.napi, eq->eq.work_done)) {
arm_bit = SET_ARM_BIT;
} else {
arm_bit = 0;
}
head = eq->head % (num_eqe << GDMA_EQE_OWNER_BITS); head = eq->head % (num_eqe << GDMA_EQE_OWNER_BITS);
mana_gd_ring_doorbell(gc, eq->gdma_dev->doorbell, eq->type, eq->id, mana_gd_ring_doorbell(gc, eq->gdma_dev->doorbell, eq->type, eq->id,
head, arm_bit); head, SET_ARM_BIT);
}
static int mana_poll(struct napi_struct *napi, int budget)
{
struct gdma_queue *eq = container_of(napi, struct gdma_queue, eq.napi);
eq->eq.work_done = 0;
eq->eq.budget = budget;
mana_gd_process_eq_events(eq);
return min(eq->eq.work_done, budget);
}
static void mana_gd_schedule_napi(void *arg)
{
struct gdma_queue *eq = arg;
struct napi_struct *napi;
napi = &eq->eq.napi;
napi_schedule_irqoff(napi);
} }
static int mana_gd_register_irq(struct gdma_queue *queue, static int mana_gd_register_irq(struct gdma_queue *queue,
const struct gdma_queue_spec *spec) const struct gdma_queue_spec *spec)
{ {
struct gdma_dev *gd = queue->gdma_dev; struct gdma_dev *gd = queue->gdma_dev;
bool is_mana = mana_gd_is_mana(gd);
struct gdma_irq_context *gic; struct gdma_irq_context *gic;
struct gdma_context *gc; struct gdma_context *gc;
struct gdma_resource *r; struct gdma_resource *r;
unsigned int msi_index; unsigned int msi_index;
unsigned long flags; unsigned long flags;
int err; struct device *dev;
int err = 0;
gc = gd->gdma_context; gc = gd->gdma_context;
r = &gc->msix_resource; r = &gc->msix_resource;
dev = gc->dev;
spin_lock_irqsave(&r->lock, flags); spin_lock_irqsave(&r->lock, flags);
msi_index = find_first_zero_bit(r->map, r->size); 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; err = -ENOSPC;
} else { } else {
bitmap_set(r->map, msi_index, 1); bitmap_set(r->map, msi_index, 1);
queue->eq.msix_index = msi_index; queue->eq.msix_index = msi_index;
err = 0;
} }
spin_unlock_irqrestore(&r->lock, flags); spin_unlock_irqrestore(&r->lock, flags);
if (err) if (err) {
return 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]; gic = &gc->irq_contexts[msi_index];
if (is_mana) {
netif_napi_add(spec->eq.ndev, &queue->eq.napi, mana_poll,
NAPI_POLL_WEIGHT);
napi_enable(&queue->eq.napi);
}
WARN_ON(gic->handler || gic->arg); WARN_ON(gic->handler || gic->arg);
gic->arg = queue; gic->arg = queue;
if (is_mana)
gic->handler = mana_gd_schedule_napi;
else
gic->handler = mana_gd_process_eq_events; gic->handler = mana_gd_process_eq_events;
return 0; return 0;
...@@ -549,11 +514,6 @@ static void mana_gd_destroy_eq(struct gdma_context *gc, bool flush_evenets, ...@@ -549,11 +514,6 @@ static void mana_gd_destroy_eq(struct gdma_context *gc, bool flush_evenets,
mana_gd_deregiser_irq(queue); mana_gd_deregiser_irq(queue);
if (mana_gd_is_mana(queue->gdma_dev)) {
napi_disable(&queue->eq.napi);
netif_napi_del(&queue->eq.napi);
}
if (queue->eq.disable_needed) if (queue->eq.disable_needed)
mana_gd_disable_queue(queue); mana_gd_disable_queue(queue);
} }
...@@ -883,6 +843,11 @@ int mana_gd_verify_vf_version(struct pci_dev *pdev) ...@@ -883,6 +843,11 @@ int mana_gd_verify_vf_version(struct pci_dev *pdev)
req.protocol_ver_min = GDMA_PROTOCOL_FIRST; req.protocol_ver_min = GDMA_PROTOCOL_FIRST;
req.protocol_ver_max = GDMA_PROTOCOL_LAST; 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); err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp), &resp);
if (err || resp.hdr.status) { if (err || resp.hdr.status) {
dev_err(gc->dev, "VfVerifyVersionOutput: %d, status=0x%x\n", dev_err(gc->dev, "VfVerifyVersionOutput: %d, status=0x%x\n",
...@@ -1128,7 +1093,7 @@ static int mana_gd_read_cqe(struct gdma_queue *cq, struct gdma_comp *comp) ...@@ -1128,7 +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; new_bits = (cq->head / num_cqe) & GDMA_CQE_OWNER_MASK;
/* Return -1 if overflow detected. */ /* Return -1 if overflow detected. */
if (owner_bits != new_bits) if (WARN_ON_ONCE(owner_bits != new_bits))
return -1; return -1;
comp->wq_num = cqe->cqe_info.wq_num; comp->wq_num = cqe->cqe_info.wq_num;
...@@ -1201,10 +1166,8 @@ static int mana_gd_setup_irqs(struct pci_dev *pdev) ...@@ -1201,10 +1166,8 @@ static int mana_gd_setup_irqs(struct pci_dev *pdev)
if (max_queues_per_port > MANA_MAX_NUM_QUEUES) if (max_queues_per_port > MANA_MAX_NUM_QUEUES)
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) */ /* 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); nvec = pci_alloc_irq_vectors(pdev, 2, max_irqs, PCI_IRQ_MSIX);
if (nvec < 0) if (nvec < 0)
...@@ -1291,6 +1254,9 @@ static int mana_gd_probe(struct pci_dev *pdev, const struct pci_device_id *ent) ...@@ -1291,6 +1254,9 @@ static int mana_gd_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
int bar = 0; int bar = 0;
int err; 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); err = pci_enable_device(pdev);
if (err) if (err)
return -ENXIO; return -ENXIO;
......
...@@ -304,7 +304,7 @@ static void mana_hwc_comp_event(void *ctx, struct gdma_queue *q_self) ...@@ -304,7 +304,7 @@ static void mana_hwc_comp_event(void *ctx, struct gdma_queue *q_self)
&comp_data); &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) static void mana_hwc_destroy_cq(struct gdma_context *gc, struct hwc_cq *hwc_cq)
......
...@@ -46,7 +46,7 @@ enum TRI_STATE { ...@@ -46,7 +46,7 @@ enum TRI_STATE {
#define EQ_SIZE (8 * PAGE_SIZE) #define EQ_SIZE (8 * PAGE_SIZE)
#define LOG2_EQ_THROTTLE 3 #define LOG2_EQ_THROTTLE 3
#define MAX_PORTS_IN_MANA_DEV 16 #define MAX_PORTS_IN_MANA_DEV 256
struct mana_stats { struct mana_stats {
u64 packets; u64 packets;
...@@ -225,6 +225,8 @@ struct mana_tx_comp_oob { ...@@ -225,6 +225,8 @@ struct mana_tx_comp_oob {
struct mana_rxq; struct mana_rxq;
#define CQE_POLLING_BUFFER 512
struct mana_cq { struct mana_cq {
struct gdma_queue *gdma_cq; struct gdma_queue *gdma_cq;
...@@ -244,8 +246,13 @@ struct mana_cq { ...@@ -244,8 +246,13 @@ struct mana_cq {
*/ */
struct mana_txq *txq; struct mana_txq *txq;
/* Pointer to a buffer which the CQ handler can copy the CQE's into. */ /* Buffer which the CQ handler can copy the CQE's into. */
struct gdma_comp *gdma_comp_buf; 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 #define GDMA_MAX_RQE_SGES 15
...@@ -315,6 +322,8 @@ struct mana_context { ...@@ -315,6 +322,8 @@ struct mana_context {
u16 num_ports; u16 num_ports;
struct mana_eq *eqs;
struct net_device *ports[MAX_PORTS_IN_MANA_DEV]; struct net_device *ports[MAX_PORTS_IN_MANA_DEV];
}; };
...@@ -324,8 +333,6 @@ struct mana_port_context { ...@@ -324,8 +333,6 @@ struct mana_port_context {
u8 mac_addr[ETH_ALEN]; u8 mac_addr[ETH_ALEN];
struct mana_eq *eqs;
enum TRI_STATE rss_state; enum TRI_STATE rss_state;
mana_handle_t default_rxobj; mana_handle_t default_rxobj;
...@@ -395,11 +402,11 @@ enum mana_command_code { ...@@ -395,11 +402,11 @@ enum mana_command_code {
struct mana_query_device_cfg_req { struct mana_query_device_cfg_req {
struct gdma_req_hdr hdr; struct gdma_req_hdr hdr;
/* Driver Capability flags */ /* MANA Nic Driver Capability flags */
u64 drv_cap_flags1; u64 mn_drv_cap_flags1;
u64 drv_cap_flags2; u64 mn_drv_cap_flags2;
u64 drv_cap_flags3; u64 mn_drv_cap_flags3;
u64 drv_cap_flags4; u64 mn_drv_cap_flags4;
u32 proto_major_ver; u32 proto_major_ver;
u32 proto_minor_ver; u32 proto_minor_ver;
...@@ -516,7 +523,7 @@ struct mana_cfg_rx_steer_resp { ...@@ -516,7 +523,7 @@ struct mana_cfg_rx_steer_resp {
struct gdma_resp_hdr hdr; struct gdma_resp_hdr hdr;
}; /* HW DATA */ }; /* HW DATA */
#define MANA_MAX_NUM_QUEUES 16 #define MANA_MAX_NUM_QUEUES 64
#define MANA_SHORT_VPORT_OFFSET_MAX ((1U << 8) - 1) #define MANA_SHORT_VPORT_OFFSET_MAX ((1U << 8) - 1)
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment