Re: [PATCH v3 02/14] soc: ti: k3: add navss ringacc driver

[Grygorii Strashko]

On 29/10/2019 10:52, Lokesh Vutla wrote:
> Hi Grygorii,
>
> [...snip..]
>
> > +
> > +static int k3_ringacc_ring_access_io(struct k3_ring *ring, void *elem,
> > +				     enum k3_ringacc_access_mode access_mode)
> > +{
> > +	void __iomem *ptr;
> > +
> > +	switch (access_mode) {
> > +	case K3_RINGACC_ACCESS_MODE_PUSH_HEAD:
> > +	case K3_RINGACC_ACCESS_MODE_POP_HEAD:
> > +		ptr = (void __iomem *)&ring->fifos->head_data;
> > +		break;
> > +	case K3_RINGACC_ACCESS_MODE_PUSH_TAIL:
> > +	case K3_RINGACC_ACCESS_MODE_POP_TAIL:
> > +		ptr = (void __iomem *)&ring->fifos->tail_data;
> > +		break;
> > +	default:
> > +		return -EINVAL;
> > +	}
> > +
> > +	ptr += k3_ringacc_ring_get_fifo_pos(ring);
> > +
> > +	switch (access_mode) {
> > +	case K3_RINGACC_ACCESS_MODE_POP_HEAD:
> > +	case K3_RINGACC_ACCESS_MODE_POP_TAIL:
> > +		dev_dbg(ring->parent->dev,
> > +			"memcpy_fromio(x): --> ptr(%p), mode:%d\n", ptr,
> > +			access_mode);
> > +		memcpy_fromio(elem, ptr, (4 << ring->elm_size));
>
> Does this work for any elem_size < 64 or any element size not aligned with 64?

Max value of elem_size is 6 as per TRM.

> IIUC, in message mode, ring element should be inserted in a single burst write
> and there is no doorbell facility. If the above conditions are not met, we are
> supposed to use proxy.
>
> In this driver, I don't see any restrictions on the ring element size for
> message mode and directly written to io. Am I missing something?

You are right and corresponding check can be added at k3_ringacc_ring_cfg() for the case
K3_RINGACC_RING_MODE_MESSAGE and no proxy

[..]

--
Best regards,
grygorii