Re: [PATCH RFC 04/12] vdpa: add vdpa simulator for block device

[Stefano Garzarella]

On Mon, Nov 16, 2020 at 12:10:19PM +0800, Jason Wang wrote:
> On 2020/11/13 下午9:47, Stefano Garzarella wrote:
> > From: Max Gurtovoy <mgurtovoy@xxxxxxxxxx>
> >
> > This will allow running vDPA for virtio block protocol.
> >
> > Signed-off-by: Max Gurtovoy <mgurtovoy@xxxxxxxxxx>
> > [sgarzare: various cleanups/fixes]
> > Signed-off-by: Stefano Garzarella <sgarzare@xxxxxxxxxx>
> > ---
> > v1:
> > - Removed unused headers
> > - Used cpu_to_vdpasim*() to store config fields
> > - Replaced 'select VDPA_SIM' with 'depends on VDPA_SIM' since selected
> >   option can not depend on other [Jason]
> > - Start with a single queue for now [Jason]
> > - Add comments to memory barriers
> > ---
> >  drivers/vdpa/vdpa_sim/vdpa_sim_blk.c | 124 +++++++++++++++++++++++++++
> >  drivers/vdpa/Kconfig                 |   9 ++
> >  drivers/vdpa/vdpa_sim/Makefile       |   1 +
> >  3 files changed, 134 insertions(+)
> >  create mode 100644 drivers/vdpa/vdpa_sim/vdpa_sim_blk.c
> >
> > diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim_blk.c b/drivers/vdpa/vdpa_sim/vdpa_sim_blk.c
> > new file mode 100644
> > index 000000000000..386dbb2f7138
> > --- /dev/null
> > +++ b/drivers/vdpa/vdpa_sim/vdpa_sim_blk.c
> > @@ -0,0 +1,124 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * VDPA simulator for block device.
> > + *
> > + * Copyright (c) 2020, Mellanox Technologies. All rights reserved.
> > + *
> > + */
> > +
> > +#include <linux/module.h>
> > +
> > +#include "vdpa_sim.h"
> > +
> > +#define VDPASIM_BLK_FEATURES	((1ULL << VIRTIO_BLK_F_SIZE_MAX) | \
> > +				 (1ULL << VIRTIO_BLK_F_SEG_MAX)  | \
> > +				 (1ULL << VIRTIO_BLK_F_BLK_SIZE) | \
> > +				 (1ULL << VIRTIO_BLK_F_TOPOLOGY) | \
> > +				 (1ULL << VIRTIO_BLK_F_MQ))
> > +
> > +#define VDPASIM_BLK_CAPACITY 0x40000
> > +#define VDPASIM_BLK_SIZE_MAX 0x1000
> > +#define VDPASIM_BLK_SEG_MAX 32
> > +#define VDPASIM_BLK_VQ_NUM 1
> > +
> > +static struct vdpasim *vdpasim_blk_dev;
> > +
> > +static void vdpasim_blk_work(struct work_struct *work)
> > +{
> > +	struct vdpasim *vdpasim = container_of(work, struct vdpasim, work);
> > +	u8 status = VIRTIO_BLK_S_OK;
> > +	int i;
> > +
> > +	spin_lock(&vdpasim->lock);
> > +
> > +	if (!(vdpasim->status & VIRTIO_CONFIG_S_DRIVER_OK))
> > +		goto out;
> > +
> > +	for (i = 0; i < VDPASIM_BLK_VQ_NUM; i++) {
> > +		struct vdpasim_virtqueue *vq = &vdpasim->vqs[i];
> > +
> > +		if (!vq->ready)
> > +			continue;
> > +
> > +		while (vringh_getdesc_iotlb(&vq->vring, &vq->iov, &vq->iov,
> > +					    &vq->head, GFP_ATOMIC) > 0) {
> > +
> > +			int write;
> > +
> > +			vq->iov.i = vq->iov.used - 1;
> > +			write = vringh_iov_push_iotlb(&vq->vring, &vq->iov, &status, 1);
> > +			if (write <= 0)
> > +				break;
> > +
> > +			/* Make sure data is wrote before advancing index */
> > +			smp_wmb();
> > +
> > +			vringh_complete_iotlb(&vq->vring, vq->head, write);
> > +
> > +			/* Make sure used is visible before rasing the interrupt. */
> > +			smp_wmb();
> > +
> > +			if (vringh_need_notify_iotlb(&vq->vring) > 0)
> > +				vringh_notify(&vq->vring);
>
>
> Do we initialize vrh->notify anywhere? And This seems duplicated with 
> the following vq->cb.
>
> I think the correct way is to initialize vrh->notify and use 
> vringh_need_notify_iotlb()/vringh_notify() instead of the vq->cb here.

Okay, so I'll set vrh->notify in the vdpasim core with a function that 
calls vq->cb() (the callback set through .set_vq_cb).

> And while at it, it's better to convert net simulator to do the same.

Sure.

Thanks,
Stefano