On Feb 20, Jakub Kicinski wrote:
> On Mon, 17 Feb 2025 14:01:16 +0100 Lorenzo Bianconi wrote:
> > +static int airoha_npu_send_msg(struct airoha_npu *npu, int func_id,
> > + void *p, int size)
> > +{
> > + u16 core = 0; /* FIXME */
> > + u32 val, offset = core << 4;
> > + dma_addr_t dma_addr;
> > + void *addr;
> > + int ret;
> > +
> > + addr = kzalloc(size, GFP_ATOMIC | GFP_DMA);
>
> You need the actual "zone DMA" memory from ISA times?
> I think that's what GFP_DMA is for. Any kmalloc'd memory
> can be DMA'ed to/from.
ack, I agree. We can drop it. I will fix it in v6.
>
> > + if (dma_set_coherent_mask(dev, 0xbfffffff))
>
> Coherent mask is not contiguous on purpose?
> Quick grep reveals no such cases at present, not sure if it works.
> Maybe add a comment, at least ?
Ack, right. I think it is wrong. I will fix it in v6.
Regards,
Lorenzo
> --
> pw-bot: cr
Attachment:
signature.asc
Description: PGP signature
