Re: [PATCH 2/2] nvme-tcp: support specifying the congestion-control

[Mingbao Sun <sunmingbao@tom.com>]

On Fri, 4 Mar 2022 17:20:32 +0100
Christoph Hellwig <hch@lst.de> wrote:

> I'll let the NVMe/TCP maintainer comment on the actual functionality,
> but:
> 
> > +			p = match_strdup(args);
> > +			if (!p) {
> > +				ret = -ENOMEM;
> > +				goto out;
> > +			}
> > +
> > +			key = tcp_ca_get_key_by_name(NULL, p,
> > &ecn_ca);
> > +			if (key == TCP_CA_UNSPEC) {
> > +				pr_err("congestion control %s not
> > found.\n",
> > +				       p);
> > +				ret = -EINVAL;
> > +				kfree(p);
> > +				goto out;
> > +			}  
> 
> We can't just call networking code from nvme-fabrics.ko

Well, actually I did have thought whether the calling of network API
here is proper. Since I did find that there is no call to APIs of
PCI/RDMA/TCP in fabrics.c.

But I hope the following could make a defense for it:

Anyway, we need to validate the tcp_congestion passed in from
user-space, right?
And it's reasonable to validate it via network API, right?

The role of nvmf_parse_options is similar to that of
drivers/nvme/target/configfs.c from the target side.
And both of them can not avoid handling specific options of the
sub-classes (e.g., NVMF_OPT_HDR_DIGEST, NVMF_OPT_TOS, NVMF_OPT_KATO).

Given the fact that the configfs.c already contains some RDMA-specific
code and has the calls to PCI-specific APIs pci_p2pdma_enable_store and
pci_p2pdma_enable_show, so I added the calling of network APIs in
configfs.c for the validation of tcp_congestion specified by the user.

So I feel this is also acceptable for nvme-fabrics.ko.