Message-ID: <PH7PR21MB326347A0E16F608B0243D9FBCE06A@PH7PR21MB3263.namprd21.prod.outlook.com>
Date:   Fri, 28 Jul 2023 21:56:57 +0000
From:   Long Li <longli@...rosoft.com>
To:     "sharmaajay@...uxonhyperv.com" <sharmaajay@...uxonhyperv.com>,
        Jason Gunthorpe <jgg@...pe.ca>,
        Leon Romanovsky <leon@...nel.org>,
        Dexuan Cui <decui@...rosoft.com>, Wei Liu <wei.liu@...nel.org>,
        "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>
CC:     "linux-rdma@...r.kernel.org" <linux-rdma@...r.kernel.org>,
        "linux-hyperv@...r.kernel.org" <linux-hyperv@...r.kernel.org>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Ajay Sharma <sharmaajay@...rosoft.com>
Subject: RE: [Patch v3 3/4] RDMA/mana_ib : Create adapter and Add error eq

> Subject: [Patch v3 3/4] RDMA/mana_ib : Create adapter and Add error eq
> 
> From: Ajay Sharma <sharmaajay@...rosoft.com>
> 
> Create adapter object as nice container for VF resources.
> Add error eq needed for adapter creation and later used for notification from
> Management SW. The management software uses this channel to send
> messages or error notifications back to the Client.
> 
> Signed-off-by: Ajay Sharma <sharmaajay@...rosoft.com>
> ---
>  drivers/infiniband/hw/mana/device.c           |  22 ++-
>  drivers/infiniband/hw/mana/main.c             |  95 ++++++++++++
>  drivers/infiniband/hw/mana/mana_ib.h          |  33 ++++
>  .../net/ethernet/microsoft/mana/gdma_main.c   | 146 ++++++++++--------
>  drivers/net/ethernet/microsoft/mana/mana_en.c |   3 +
>  include/net/mana/gdma.h                       |  13 +-
>  6 files changed, 242 insertions(+), 70 deletions(-)
> 
> diff --git a/drivers/infiniband/hw/mana/device.c
> b/drivers/infiniband/hw/mana/device.c
> index ea4c8c8fc10d..4077e440657a 100644
> --- a/drivers/infiniband/hw/mana/device.c
> +++ b/drivers/infiniband/hw/mana/device.c
> @@ -68,7 +68,7 @@ static int mana_ib_probe(struct auxiliary_device *adev,
>  	ibdev_dbg(&mib_dev->ib_dev, "mdev=%p id=%d num_ports=%d\n",
> mdev,
>  		  mdev->dev_id.as_uint32, mib_dev->ib_dev.phys_port_cnt);
> 
> -	mib_dev->gdma_dev = mdev;
> +	mib_dev->gc = mdev->gdma_context;
>  	mib_dev->ib_dev.node_type = RDMA_NODE_IB_CA;
> 
>  	/*
> @@ -85,15 +85,31 @@ static int mana_ib_probe(struct auxiliary_device
> *adev,
>  		goto free_ib_device;
>  	}
> 
> +	ret = mana_ib_create_error_eq(mib_dev);
> +	if (ret) {
> +		ibdev_err(&mib_dev->ib_dev, "Failed to allocate err eq");
> +		goto deregister_device;
> +	}
> +
> +	ret = mana_ib_create_adapter(mib_dev);
> +	if (ret) {
> +		ibdev_err(&mib_dev->ib_dev, "Failed to create adapter");
> +		goto free_error_eq;
> +	}
> +
>  	ret = ib_register_device(&mib_dev->ib_dev, "mana_%d",
>  				 mdev->gdma_context->dev);
>  	if (ret)
> -		goto deregister_device;
> +		goto destroy_adapter;
> 
>  	dev_set_drvdata(&adev->dev, mib_dev);
> 
>  	return 0;
> 
> +destroy_adapter:
> +	mana_ib_destroy_adapter(mib_dev);
> +free_error_eq:
> +	mana_gd_destroy_queue(mib_dev->gc, mib_dev->fatal_err_eq);
>  deregister_device:
>  	mana_gd_deregister_device(&mib_dev->gc->mana_ib);
>  free_ib_device:
> @@ -105,6 +121,8 @@ static void mana_ib_remove(struct auxiliary_device
> *adev)  {
>  	struct mana_ib_dev *mib_dev = dev_get_drvdata(&adev->dev);
> 
> +	mana_gd_destroy_queue(mib_dev->gc, mib_dev->fatal_err_eq);
> +	mana_ib_destroy_adapter(mib_dev);
>  	mana_gd_deregister_device(&mib_dev->gc->mana_ib);
>  	ib_unregister_device(&mib_dev->ib_dev);
>  	ib_dealloc_device(&mib_dev->ib_dev);
> diff --git a/drivers/infiniband/hw/mana/main.c
> b/drivers/infiniband/hw/mana/main.c
> index 2c4e3c496644..1b1a8670d0fa 100644
> --- a/drivers/infiniband/hw/mana/main.c
> +++ b/drivers/infiniband/hw/mana/main.c
> @@ -504,3 +504,98 @@ int mana_ib_query_gid(struct ib_device *ibdev, u32
> port, int index,  void mana_ib_disassociate_ucontext(struct ib_ucontext
> *ibcontext)  {  }
> +
> +int mana_ib_destroy_adapter(struct mana_ib_dev *mib_dev) {
> +	struct mana_ib_destroy_adapter_resp resp = {};
> +	struct mana_ib_destroy_adapter_req req = {};
> +	struct gdma_context *gc;
> +	int err;
> +
> +	gc = mib_dev->gc;
> +
> +	mana_gd_init_req_hdr(&req.hdr, MANA_IB_DESTROY_ADAPTER,
> sizeof(req),
> +			     sizeof(resp));
> +	req.adapter = mib_dev->adapter_handle;
> +	req.hdr.dev_id = gc->mana_ib.dev_id;
> +
> +	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp),
> +&resp);
> +
> +	if (err) {
> +		ibdev_err(&mib_dev->ib_dev, "Failed to destroy adapter
> err %d", err);
> +		return err;
> +	}
> +
> +	return 0;
> +}
> +
> +int mana_ib_create_adapter(struct mana_ib_dev *mib_dev) {
> +	struct mana_ib_create_adapter_resp resp = {};
> +	struct mana_ib_create_adapter_req req = {};
> +	struct gdma_context *gc;
> +	int err;
> +
> +	gc = mib_dev->gc;
> +
> +	mana_gd_init_req_hdr(&req.hdr, MANA_IB_CREATE_ADAPTER,
> sizeof(req),
> +			     sizeof(resp));
> +	req.notify_eq_id = mib_dev->fatal_err_eq->id;
> +	req.hdr.dev_id = gc->mana_ib.dev_id;
> +
> +	err = mana_gd_send_request(gc, sizeof(req), &req, sizeof(resp),
> +&resp);
> +
> +	if (err) {
> +		ibdev_err(&mib_dev->ib_dev, "Failed to create adapter
> err %d",
> +			  err);
> +		return err;
> +	}
> +
> +	mib_dev->adapter_handle = resp.adapter;
> +
> +	return 0;
> +}
> +
> +static void mana_ib_soc_event_handler(void *ctx, struct gdma_queue
> *queue,
> +				      struct gdma_event *event)
> +{
> +	struct mana_ib_dev *mib_dev = (struct mana_ib_dev *)ctx;
> +
> +	switch (event->type) {
> +	case GDMA_EQE_SOC_EVENT_NOTIFICATION:
> +		ibdev_info(&mib_dev->ib_dev, "Received SOC Notification");
> +		break;

Should we do something with the event?




> +	case GDMA_EQE_SOC_EVENT_TEST:
> +		ibdev_info(&mib_dev->ib_dev, "Received SoC Test");
> +		break;
> +	default:
> +		ibdev_dbg(&mib_dev->ib_dev, "Received unsolicited evt %d",
> +			  event->type);
> +	}
> +}
> +
> +int mana_ib_create_error_eq(struct mana_ib_dev *mib_dev) {
> +	struct gdma_queue_spec spec = {};
> +	int err;
> +
> +	spec.type = GDMA_EQ;
> +	spec.monitor_avl_buf = false;
> +	spec.queue_size = EQ_SIZE;
> +	spec.eq.callback = mana_ib_soc_event_handler;
> +	spec.eq.context = mib_dev;
> +	spec.eq.log2_throttle_limit = LOG2_EQ_THROTTLE;
> +	spec.eq.msix_allocated = true;
> +	spec.eq.msix_index = 0;
> +	spec.doorbell = mib_dev->gc->mana_ib.doorbell;
> +	spec.pdid = mib_dev->gc->mana_ib.pdid;
> +
> +	err = mana_gd_create_mana_eq(&mib_dev->gc->mana_ib, &spec,
> +				     &mib_dev->fatal_err_eq);
> +	if (err)
> +		return err;
> +
> +	mib_dev->fatal_err_eq->eq.disable_needed = true;
> +
> +	return 0;
> +}
> diff --git a/drivers/infiniband/hw/mana/mana_ib.h
> b/drivers/infiniband/hw/mana/mana_ib.h
> index 3a2ba6b96f15..8a652bccd978 100644
> --- a/drivers/infiniband/hw/mana/mana_ib.h
> +++ b/drivers/infiniband/hw/mana/mana_ib.h
> @@ -31,6 +31,8 @@ struct mana_ib_dev {
>  	struct ib_device ib_dev;
>  	struct gdma_dev *gdma_dev;
>  	struct gdma_context *gc;
> +	struct gdma_queue *fatal_err_eq;
> +	mana_handle_t adapter_handle;
>  };
> 
>  struct mana_ib_wq {
> @@ -93,6 +95,31 @@ struct mana_ib_rwq_ind_table {
>  	struct ib_rwq_ind_table ib_ind_table;
>  };
> 
> +enum mana_ib_command_code {
> +	MANA_IB_CREATE_ADAPTER  = 0x30002,
> +	MANA_IB_DESTROY_ADAPTER = 0x30003,
> +};
> +
> +struct mana_ib_create_adapter_req {
> +	struct gdma_req_hdr hdr;
> +	u32 notify_eq_id;
> +	u32 reserved;
> +}; /*HW Data */
> +
> +struct mana_ib_create_adapter_resp {
> +	struct gdma_resp_hdr hdr;
> +	mana_handle_t adapter;
> +}; /* HW Data */
> +
> +struct mana_ib_destroy_adapter_req {
> +	struct gdma_req_hdr hdr;
> +	mana_handle_t adapter;
> +}; /*HW Data */
> +
> +struct mana_ib_destroy_adapter_resp {
> +	struct gdma_resp_hdr hdr;
> +}; /* HW Data */
> +
>  int mana_ib_gd_create_dma_region(struct mana_ib_dev *mib_dev,
>  				 struct ib_umem *umem,
>  				 mana_handle_t *gdma_region);
> @@ -161,4 +188,10 @@ int mana_ib_query_gid(struct ib_device *ibdev, u32
> port, int index,
> 
>  void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext);
> 
> +int mana_ib_create_error_eq(struct mana_ib_dev *mib_dev);
> +
> +int mana_ib_create_adapter(struct mana_ib_dev *mib_dev);
> +
> +int mana_ib_destroy_adapter(struct mana_ib_dev *mib_dev);
> +
>  #endif
> diff --git a/drivers/net/ethernet/microsoft/mana/gdma_main.c
> b/drivers/net/ethernet/microsoft/mana/gdma_main.c
> index 9fa7a2d6c2b2..55e194c9d84e 100644
> --- a/drivers/net/ethernet/microsoft/mana/gdma_main.c
> +++ b/drivers/net/ethernet/microsoft/mana/gdma_main.c
> @@ -185,7 +185,8 @@ void mana_gd_free_memory(struct gdma_mem_info
> *gmi)  }
> 
>  static int mana_gd_create_hw_eq(struct gdma_context *gc,
> -				struct gdma_queue *queue)
> +				struct gdma_queue *queue,
> +				u32 doorbell, u32 pdid)
>  {
>  	struct gdma_create_queue_resp resp = {};
>  	struct gdma_create_queue_req req = {}; @@ -199,8 +200,8 @@ static
> int mana_gd_create_hw_eq(struct gdma_context *gc,
> 
>  	req.hdr.dev_id = queue->gdma_dev->dev_id;
>  	req.type = queue->type;
> -	req.pdid = queue->gdma_dev->pdid;
> -	req.doolbell_id = queue->gdma_dev->doorbell;
> +	req.pdid = pdid;
> +	req.doolbell_id = doorbell;
>  	req.gdma_region = queue->mem_info.dma_region_handle;
>  	req.queue_size = queue->queue_size;
>  	req.log2_throttle_limit = queue->eq.log2_throttle_limit; @@ -371,53
> +372,51 @@ static void mana_gd_process_eqe(struct gdma_queue *eq)
>  	}
>  }
> 
> -static void mana_gd_process_eq_events(void *arg)
> +static void mana_gd_process_eq_events(struct list_head *eq_list)
>  {
>  	u32 owner_bits, new_bits, old_bits;
>  	union gdma_eqe_info eqe_info;
>  	struct gdma_eqe *eq_eqe_ptr;
> -	struct gdma_queue *eq = arg;
> +	struct gdma_queue *eq;
>  	struct gdma_context *gc;
>  	struct gdma_eqe *eqe;
>  	u32 head, num_eqe;
>  	int i;
> 
> -	gc = eq->gdma_dev->gdma_context;
> -
> -	num_eqe = eq->queue_size / GDMA_EQE_SIZE;
> -	eq_eqe_ptr = eq->queue_mem_ptr;
> -
> -	/* Process up to 5 EQEs at a time, and update the HW head. */
> -	for (i = 0; i < 5; i++) {
> -		eqe = &eq_eqe_ptr[eq->head % num_eqe];
> -		eqe_info.as_uint32 = eqe->eqe_info;
> -		owner_bits = eqe_info.owner_bits;
> -
> -		old_bits = (eq->head / num_eqe - 1) &
> GDMA_EQE_OWNER_MASK;
> -		/* No more entries */
> -		if (owner_bits == old_bits)
> -			break;
> -
> -		new_bits = (eq->head / num_eqe) &
> GDMA_EQE_OWNER_MASK;
> -		if (owner_bits != new_bits) {
> -			dev_err(gc->dev, "EQ %d: overflow detected\n", eq-
> >id);
> -			break;
> +	list_for_each_entry_rcu(eq, eq_list, entry) {
> +		gc = eq->gdma_dev->gdma_context;
> +
> +		num_eqe = eq->queue_size / GDMA_EQE_SIZE;
> +		eq_eqe_ptr = eq->queue_mem_ptr;
> +		/* Process up to 5 EQEs at a time, and update the HW head. */
> +		for (i = 0; i < 5; i++) {
> +			eqe = &eq_eqe_ptr[eq->head % num_eqe];
> +			eqe_info.as_uint32 = eqe->eqe_info;
> +			owner_bits = eqe_info.owner_bits;
> +
> +			old_bits = (eq->head / num_eqe - 1) &
> GDMA_EQE_OWNER_MASK;
> +			/* No more entries */
> +			if (owner_bits == old_bits)
> +				break;
> +
> +			new_bits = (eq->head / num_eqe) &
> GDMA_EQE_OWNER_MASK;
> +			if (owner_bits != new_bits) {
> +				dev_err(gc->dev, "EQ %d: overflow
> detected\n",
> +					eq->id);
> +				break;
> +			}
> +			/* Per GDMA spec, rmb is necessary after checking
> owner_bits, before
> +			 * reading eqe.
> +			 */
> +			rmb();
> +			mana_gd_process_eqe(eq);
> +			eq->head++;
>  		}
> 
> -		/* Per GDMA spec, rmb is necessary after checking
> owner_bits, before
> -		 * reading eqe.
> -		 */
> -		rmb();
> -
> -		mana_gd_process_eqe(eq);
> -
> -		eq->head++;
> +		head = eq->head % (num_eqe << GDMA_EQE_OWNER_BITS);
> +		mana_gd_ring_doorbell(gc, eq->gdma_dev->doorbell, eq-
> >type,
> +				      eq->id, head, SET_ARM_BIT);
>  	}
> -
> -	head = eq->head % (num_eqe << GDMA_EQE_OWNER_BITS);
> -
> -	mana_gd_ring_doorbell(gc, eq->gdma_dev->doorbell, eq->type, eq-
> >id,
> -			      head, SET_ARM_BIT);
>  }
> 
>  static int mana_gd_register_irq(struct gdma_queue *queue, @@ -435,44
> +434,47 @@ static int mana_gd_register_irq(struct gdma_queue *queue,
>  	gc = gd->gdma_context;
>  	r = &gc->msix_resource;
>  	dev = gc->dev;
> +	msi_index = spec->eq.msix_index;
> 
>  	spin_lock_irqsave(&r->lock, flags);
> 
> -	msi_index = find_first_zero_bit(r->map, 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;
> -	}
> -
> -	spin_unlock_irqrestore(&r->lock, flags);
> +	if (!spec->eq.msix_allocated) {
> +		msi_index = find_first_zero_bit(r->map, r->size);
> 
> -	if (err) {
> -		dev_err(dev, "Register IRQ err:%d, msi:%u rsize:%u,
> nMSI:%u",
> -			err, msi_index, r->size, gc->num_msix_usable);
> +		if (msi_index >= r->size ||
> +		    msi_index >= gc->num_msix_usable)
> +			err = -ENOSPC;
> +		else
> +			bitmap_set(r->map, msi_index, 1);
> 
> -		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);
> +				goto out;
> +		}
>  	}
> 
> +	queue->eq.msix_index = msi_index;
>  	gic = &gc->irq_contexts[msi_index];
> 
> -	WARN_ON(gic->handler || gic->arg);
> -
> -	gic->arg = queue;
> +	list_add_rcu(&queue->entry, &gic->eq_list);
> 
>  	gic->handler = mana_gd_process_eq_events;
> 
> -	return 0;
> +out:
> +	spin_unlock_irqrestore(&r->lock, flags);
> +	return err;
>  }
> 
> -static void mana_gd_deregiser_irq(struct gdma_queue *queue)
> +static void mana_gd_deregister_irq(struct gdma_queue *queue)
>  {
>  	struct gdma_dev *gd = queue->gdma_dev;
>  	struct gdma_irq_context *gic;
>  	struct gdma_context *gc;
>  	struct gdma_resource *r;
>  	unsigned int msix_index;
> +	struct list_head *p, *n;
> +	struct gdma_queue *eq;
>  	unsigned long flags;
> 
>  	gc = gd->gdma_context;
> @@ -483,14 +485,23 @@ static void mana_gd_deregiser_irq(struct
> gdma_queue *queue)
>  	if (WARN_ON(msix_index >= gc->num_msix_usable))
>  		return;
> 
> +	spin_lock_irqsave(&r->lock, flags);
> +
>  	gic = &gc->irq_contexts[msix_index];
> -	gic->handler = NULL;
> -	gic->arg = NULL;
> +	list_for_each_safe(p, n, &gic->eq_list) {
> +		eq = list_entry(p, struct gdma_queue, entry);
> +		if (queue == eq) {
> +			list_del(&eq->entry);

The previous code used list_for_each_entry_rcu() for iterating eq, need to add rcu_synchronize()?




> +			break;
> +		}
> +	}
> 

Powered by blists - more mailing lists