Re: [PATCH rdma-next v2 07/11] RDMA/erdma: Add verbs implementation

[Jason Gunthorpe <jgg@ziepe.ca>]

On Mon, Jan 17, 2022 at 04:48:24PM +0800, Cheng Xu wrote:

> +	entry = to_emmap(rdma_entry);
> +
> +	switch (entry->mmap_flag) {
> +	case ERDMA_MMAP_IO_NC:
> +		/* map doorbell. */
> +		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);

This should be pgprot_device()

> +		err = io_remap_pfn_range(vma, vma->vm_start, PFN_DOWN(entry->address),
> +			PAGE_SIZE, vma->vm_page_prot);
> +		break;
> +	default:
> +		pr_err("mmap failed, mmap_flag = %d\n", entry->mmap_flag);

Audit this whole driver for prints:
 1) Do not ever print on user spcae triggered paths, like this
 2) Always use ibdev_dbg/err/etc

> +void erdma_disassociate_ucontext(struct ib_ucontext *ibcontext)
> +{
> +}

Does the driver really fully support device disassociation, including
support to recover from itin rdma-core? Don't define this function otherwise.

Jason