RE: [PATCH v2 6/8] vdpa/ifc: support dynamic enable/disable queue

["Pei, Andy" <andy.pei@intel.com>]

Hi Chenbo, 

See my reply inline.

> -----Original Message-----
> From: Huang, Wei <wei.huang@intel.com>
> Sent: Wednesday, September 14, 2022 11:04 AM
> To: Xia, Chenbo <Chenbo.Xia@intel.com>; Pei, Andy <andy.pei@intel.com>;
> dev@dpdk.org
> Cc: Xu, Rosen <rosen.xu@intel.com>; Cao, Gang <gang.cao@intel.com>;
> maxime.coquelin@redhat.com; Huang Wei <wei_huang@intel.com>
> Subject: RE: [PATCH v2 6/8] vdpa/ifc: support dynamic enable/disable queue
> 
> 
> 
> > -----Original Message-----
> > From: Xia, Chenbo <chenbo.xia@intel.com>
> > Sent: Wednesday, September 14, 2022 10:23
> > To: Pei, Andy <andy.pei@intel.com>; dev@dpdk.org
> > Cc: Xu, Rosen <rosen.xu@intel.com>; Huang, Wei <wei.huang@intel.com>;
> > Cao, Gang <gang.cao@intel.com>; maxime.coquelin@redhat.com; Huang Wei
> > <wei_huang@intel.com>
> > Subject: RE: [PATCH v2 6/8] vdpa/ifc: support dynamic enable/disable
> > queue
> >
> > > -----Original Message-----
> > > From: Pei, Andy <andy.pei@intel.com>
> > > Sent: Thursday, September 8, 2022 1:54 PM
> > > To: dev@dpdk.org
> > > Cc: Xia, Chenbo <chenbo.xia@intel.com>; Xu, Rosen
> > > <rosen.xu@intel.com>; Huang, Wei <wei.huang@intel.com>; Cao, Gang
> > > <gang.cao@intel.com>; maxime.coquelin@redhat.com; Huang Wei
> > > <wei_huang@intel.com>
> > > Subject: [PATCH v2 6/8] vdpa/ifc: support dynamic enable/disable
> > > queue
> > >
> > > From: Huang Wei <wei_huang@intel.com>
> > >
> > > Support dynamic enable or disable queue.
> > > For front end, like QEMU, user can use ethtool to configurate queue.
> > > For example, "ethtool -L eth0 combined 3" to enable 3 queues pairs.
> > >
> > > Signed-off-by: Huang Wei <wei_huang@intel.com>
> > > Signed-off-by: Andy Pei <andy.pei@intel.com>
> > > ---
> > >  drivers/vdpa/ifc/base/ifcvf.c       | 101
> > > ++++++++++++++++++++++++++++++++++++
> > >  drivers/vdpa/ifc/base/ifcvf.h       |   6 +++
> > >  drivers/vdpa/ifc/base/ifcvf_osdep.h |   1 +
> > >  drivers/vdpa/ifc/ifcvf_vdpa.c       |  93 +++++++++++++++++++++++++++----
> > > --
> > >  4 files changed, 186 insertions(+), 15 deletions(-)
> > >
> > > diff --git a/drivers/vdpa/ifc/base/ifcvf.c
> > > b/drivers/vdpa/ifc/base/ifcvf.c index 4875ea1..34f32f8 100644
> > > --- a/drivers/vdpa/ifc/base/ifcvf.c
> > > +++ b/drivers/vdpa/ifc/base/ifcvf.c
> > > @@ -230,6 +230,107 @@
> > >  	}
> > >  }
> > >
> > > +int
> > > +ifcvf_enable_vring_hw(struct ifcvf_hw *hw, int i) {
> > > +	struct ifcvf_pci_common_cfg *cfg;
> > > +	u8 *lm_cfg;
> > > +	u16 notify_off;
> > > +	int msix_vector;
> > > +
> > > +	if (!hw || (i >= (int)hw->nr_vring))
> > > +		return -1;
> >
> > Seems HW will always be not NULL
> As a external function, we should not assume the input argument is always valid.
> >
> > > +
> > > +	cfg = hw->common_cfg;
> > > +	if (!cfg) {
> > > +		ERROUT("common_cfg in HW is NULL.\n");
> >
> > I am thinking why you introduce this new log? Why not just use DRV_LOG
> > that is already defined?
> Because below type of log macros are already defined and used in original code,
> I just follow its tradition.
> #define WARNINGOUT(S, args...)  RTE_LOG(WARNING, PMD, S, ##args)
> #define DEBUGOUT(S, args...)    RTE_LOG(DEBUG, PMD, S, ##args)
> #define ERROUT(S, args...)      RTE_LOG(ERR, PMD, S, ##args)
> >
> > > +		return -1;
> > > +	}
> > > +
> > > +	ifcvf_enable_multiqueue(hw);
> > > +
> > > +	IFCVF_WRITE_REG16(i, &cfg->queue_select);
> > > +	msix_vector = IFCVF_READ_REG16(&cfg->queue_msix_vector);
> > > +	if (msix_vector != (i + 1)) {
> > > +		IFCVF_WRITE_REG16(i + 1, &cfg->queue_msix_vector);
> > > +		msix_vector = IFCVF_READ_REG16(&cfg->queue_msix_vector);
> > > +		if (msix_vector == IFCVF_MSI_NO_VECTOR) {
> > > +			ERROUT("queue %u, msix vec alloc failed\n", i);
> > > +			return -1;
> > > +		}
> > > +	}
> > > +
> > > +	io_write64_twopart(hw->vring[i].desc, &cfg->queue_desc_lo,
> > > +			&cfg->queue_desc_hi);
> > > +	io_write64_twopart(hw->vring[i].avail, &cfg->queue_avail_lo,
> > > +			&cfg->queue_avail_hi);
> > > +	io_write64_twopart(hw->vring[i].used, &cfg->queue_used_lo,
> > > +			&cfg->queue_used_hi);
> > > +	IFCVF_WRITE_REG16(hw->vring[i].size, &cfg->queue_size);
> > > +
> > > +	lm_cfg = hw->lm_cfg;
> > > +	if (lm_cfg) {
> > > +		if (hw->device_type == IFCVF_BLK)
> > > +			*(u32 *)(lm_cfg + IFCVF_LM_RING_STATE_OFFSET +
> > > +				i * IFCVF_LM_CFG_SIZE) =
> > > +				(u32)hw->vring[i].last_avail_idx |
> > > +				((u32)hw->vring[i].last_used_idx << 16);
> > > +		else
> > > +			*(u32 *)(lm_cfg + IFCVF_LM_RING_STATE_OFFSET +
> > > +				(i / 2) * IFCVF_LM_CFG_SIZE +
> > > +				(i % 2) * 4) =
> > > +				(u32)hw->vring[i].last_avail_idx |
> > > +				((u32)hw->vring[i].last_used_idx << 16);
> > > +	}
> >
> > So the register layout is different for blk and net?
> That's sure.
For the register layout differences, the story is as follow:
When I add support for blk device, I tried to re-use the existing code.
However, the BAR4 layout of blk HW is different.
I keep code for net device unchanged, and add code for blk device.
> >
> > > +
> > > +	notify_off = IFCVF_READ_REG16(&cfg->queue_notify_off);
> > > +	hw->notify_addr[i] = (void *)((u8 *)hw->notify_base +
> > > +			notify_off * hw->notify_off_multiplier);
> > > +	IFCVF_WRITE_REG16(1, &cfg->queue_enable);
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +void
> > > +ifcvf_disable_vring_hw(struct ifcvf_hw *hw, int i) {
> > > +	struct ifcvf_pci_common_cfg *cfg;
> > > +	u32 ring_state;
> > > +	u8 *lm_cfg;
> > > +
> > > +	if (!hw || (i >= (int)hw->nr_vring))
> > > +		return;
> > > +
> > > +	cfg = hw->common_cfg;
> > > +	if (!cfg) {
> > > +		ERROUT("common_cfg in HW is NULL.\n");
> > > +		return;
> > > +	}
> > > +
> > > +	IFCVF_WRITE_REG16(i, &cfg->queue_select);
> > > +	IFCVF_WRITE_REG16(0, &cfg->queue_enable);
> > > +
> > > +	lm_cfg = hw->lm_cfg;
> > > +	if (lm_cfg) {
> > > +		if (hw->device_type == IFCVF_BLK)
> > > +			ring_state = *(u32 *)(lm_cfg +
> > > +					IFCVF_LM_RING_STATE_OFFSET +
> > > +					i * IFCVF_LM_CFG_SIZE);
> > > +		else
> > > +			ring_state = *(u32 *)(lm_cfg +
> > > +					IFCVF_LM_RING_STATE_OFFSET +
> > > +					(i / 2) * IFCVF_LM_CFG_SIZE +
> > > +					(i % 2) * 4);
> > > +
> > > +		if (hw->device_type == IFCVF_BLK)
> > > +			hw->vring[i].last_avail_idx =
> > > +				(u16)(ring_state & IFCVF_16_BIT_MASK);
> > > +		else
> > > +			hw->vring[i].last_avail_idx = (u16)(ring_state >> 16);
> >
> > Above two if-else should be combined.
> It's a good suggestion.
> >
> > Thanks,
> > Chenbo
> >
> > > +		hw->vring[i].last_used_idx = (u16)(ring_state >> 16);
> > > +	}
> > > +}
> > > +
> > >  STATIC int
> > >  ifcvf_hw_enable(struct ifcvf_hw *hw)  { diff --git
> > > a/drivers/vdpa/ifc/base/ifcvf.h b/drivers/vdpa/ifc/base/ifcvf.h
> > > index c17bf2a..e67d4e8 100644
> > > --- a/drivers/vdpa/ifc/base/ifcvf.h
> > > +++ b/drivers/vdpa/ifc/base/ifcvf.h
> > > @@ -164,6 +164,12 @@ struct ifcvf_hw {  ifcvf_get_features(struct
> > > ifcvf_hw *hw);
> > >
> > >  int
> > > +ifcvf_enable_vring_hw(struct ifcvf_hw *hw, int i);
> > > +
> > > +void
> > > +ifcvf_disable_vring_hw(struct ifcvf_hw *hw, int i);
> > > +
> > > +int
> > >  ifcvf_start_hw(struct ifcvf_hw *hw);
> > >
> > >  void
> > > diff --git a/drivers/vdpa/ifc/base/ifcvf_osdep.h
> > > b/drivers/vdpa/ifc/base/ifcvf_osdep.h
> > > index 3d56769..4a1bfec 100644
> > > --- a/drivers/vdpa/ifc/base/ifcvf_osdep.h
> > > +++ b/drivers/vdpa/ifc/base/ifcvf_osdep.h
> > > @@ -16,6 +16,7 @@
> > >
> > >  #define WARNINGOUT(S, args...)  RTE_LOG(WARNING, PMD, S, ##args)
> > >  #define DEBUGOUT(S, args...)    RTE_LOG(DEBUG, PMD, S, ##args)
> > > +#define ERROUT(S, args...)      RTE_LOG(ERR, PMD, S, ##args)
> > >  #define STATIC                  static
> > >
> > >  #define msec_delay(x)	rte_delay_us_sleep(1000 * (x))
> > > diff --git a/drivers/vdpa/ifc/ifcvf_vdpa.c
> > > b/drivers/vdpa/ifc/ifcvf_vdpa.c index 48f1a89..16fd0fd 100644
> > > --- a/drivers/vdpa/ifc/ifcvf_vdpa.c
> > > +++ b/drivers/vdpa/ifc/ifcvf_vdpa.c
> > > @@ -1288,13 +1288,59 @@ struct rte_vdpa_dev_info {  }
> > >
> > >  static int
> > > +ifcvf_config_vring(struct ifcvf_internal *internal, int vring) {
> > > +	struct ifcvf_hw *hw = &internal->hw;
> > > +	int vid = internal->vid;
> > > +	struct rte_vhost_vring vq;
> > > +	uint64_t gpa;
> > > +
> > > +	if (hw->vring[vring].enable) {
> > > +		rte_vhost_get_vhost_vring(vid, vring, &vq);
> > > +		gpa = hva_to_gpa(vid, (uint64_t)(uintptr_t)vq.desc);
> > > +		if (gpa == 0) {
> > > +			DRV_LOG(ERR, "Fail to get GPA for descriptor ring.");
> > > +			return -1;
> > > +		}
> > > +		hw->vring[vring].desc = gpa;
> > > +
> > > +		gpa = hva_to_gpa(vid, (uint64_t)(uintptr_t)vq.avail);
> > > +		if (gpa == 0) {
> > > +			DRV_LOG(ERR, "Fail to get GPA for available ring.");
> > > +			return -1;
> > > +		}
> > > +		hw->vring[vring].avail = gpa;
> > > +
> > > +		gpa = hva_to_gpa(vid, (uint64_t)(uintptr_t)vq.used);
> > > +		if (gpa == 0) {
> > > +			DRV_LOG(ERR, "Fail to get GPA for used ring.");
> > > +			return -1;
> > > +		}
> > > +		hw->vring[vring].used = gpa;
> > > +
> > > +		hw->vring[vring].size = vq.size;
> > > +		rte_vhost_get_vring_base(vid, vring,
> > > +				&hw->vring[vring].last_avail_idx,
> > > +				&hw->vring[vring].last_used_idx);
> > > +		ifcvf_enable_vring_hw(&internal->hw, vring);
> > > +	} else {
> > > +		ifcvf_disable_vring_hw(&internal->hw, vring);
> > > +		rte_vhost_set_vring_base(vid, vring,
> > > +				hw->vring[vring].last_avail_idx,
> > > +				hw->vring[vring].last_used_idx);
> > > +	}
> > > +
> > > +	return 0;
> > > +}
> > > +
> > > +static int
> > >  ifcvf_set_vring_state(int vid, int vring, int state)  {
> > >  	struct rte_vdpa_device *vdev;
> > >  	struct internal_list *list;
> > >  	struct ifcvf_internal *internal;
> > >  	struct ifcvf_hw *hw;
> > > -	struct ifcvf_pci_common_cfg *cfg;
> > > +	bool enable = !!state;
> > >  	int ret = 0;
> > >
> > >  	vdev = rte_vhost_get_vdpa_device(vid); @@ -1304,6 +1350,9 @@
> > struct
> > > rte_vdpa_dev_info {
> > >  		return -1;
> > >  	}
> > >
> > > +	DRV_LOG(INFO, "%s queue %d of vDPA device %s",
> > > +		enable ? "enable" : "disable", vring, vdev->device->name);
> > > +
> > >  	internal = list->internal;
> > >  	if (vring < 0 || vring >= internal->max_queues * 2) {
> > >  		DRV_LOG(ERR, "Vring index %d not correct", vring); @@ -
> > 1311,27
> > > +1360,41 @@ struct rte_vdpa_dev_info {
> > >  	}
> > >
> > >  	hw = &internal->hw;
> > > +	hw->vring[vring].enable = enable;
> > > +
> > >  	if (!internal->configured)
> > > -		goto exit;
> > > +		return 0;
> > >
> > > -	cfg = hw->common_cfg;
> > > -	IFCVF_WRITE_REG16(vring, &cfg->queue_select);
> > > -	IFCVF_WRITE_REG16(!!state, &cfg->queue_enable);
> > > +	unset_notify_relay(internal);
> > >
> > > -	if (!state && hw->vring[vring].enable) {
> > > -		ret = vdpa_disable_vfio_intr(internal);
> > > -		if (ret)
> > > -			return ret;
> > > +	ret = vdpa_enable_vfio_intr(internal, false);
> > > +	if (ret) {
> > > +		DRV_LOG(ERR, "failed to set vfio interrupt of vDPA device %s",
> > > +			vdev->device->name);
> > > +		return ret;
> > >  	}
> > >
> > > -	if (state && !hw->vring[vring].enable) {
> > > -		ret = vdpa_enable_vfio_intr(internal, false);
> > > -		if (ret)
> > > -			return ret;
> > > +	ret = ifcvf_config_vring(internal, vring);
> > > +	if (ret) {
> > > +		DRV_LOG(ERR, "failed to configure queue %d of vDPA
> > device %s",
> > > +			vring, vdev->device->name);
> > > +		return ret;
> > > +	}
> > > +
> > > +	ret = setup_notify_relay(internal);
> > > +	if (ret) {
> > > +		DRV_LOG(ERR, "failed to setup notify relay of vDPA device %s",
> > > +			vdev->device->name);
> > > +		return ret;
> > > +	}
> > > +
> > > +	ret = rte_vhost_host_notifier_ctrl(vid, vring, enable);
> > > +	if (ret) {
> > > +		DRV_LOG(ERR, "vDPA device %s queue %d host notifier ctrl fail",
> > > +			vdev->device->name, vring);
> > > +		return ret;
> > >  	}
> > >
> > > -exit:
> > > -	hw->vring[vring].enable = !!state;
> > >  	return 0;
> > >  }
> > >
> > > --
> > > 1.8.3.1