Re: [PATCH 1/1] Add 'Transport Interface' (triface) option. This can be used to specify the IP interface to use for the connection. The driver uses that to set SO_BINDTODEVICE on the socket before connecting.

From: Hannes Reinecke <hare@suse.de>
To: "Belanger, Martin" <Martin.Belanger@dell.com>,
	Martin Belanger <nitram_67@hotmail.com>,
	"linux-nvme@lists.infradead.org" <linux-nvme@lists.infradead.org>
Cc: "kbusch@kernel.org" <kbusch@kernel.org>,
	"axboe@fb.com" <axboe@fb.com>, "hch@lst.de" <hch@lst.de>,
	"sagi@grimberg.me" <sagi@grimberg.me>
Subject: Re: [PATCH 1/1] Add 'Transport Interface' (triface) option. This can be used to specify the IP interface to use for the connection. The driver uses that to set SO_BINDTODEVICE on the socket before connecting.
Date: Tue, 4 May 2021 15:25:41 +0200	[thread overview]
Message-ID: <cc8805b2-63eb-93ac-6505-66305a73cb3c@suse.de> (raw)
In-Reply-To: <SJ0PR19MB454486D08CA07BCE274966E6F25B9@SJ0PR19MB4544.namprd19.prod.outlook.com>

On 5/3/21 6:59 PM, Belanger, Martin wrote:
> Hi Hannes,
> 
> I just noticed there were in-line comments. to answer you questions:
> 

This is the recommended style when working with linux patches; please do 
not top-post.

> Q1) Why not simply 'host_iface' ? 'triface' is a bit awkward.
> A1) I used TRIFACE to keep consistency with all other transport options: traddr, trsvcid, host_traddr.
> I will rename to host_iface at your suggestion.
> 
> Q2) Is this valid for all transports? I guess it would only work for 'tcp', and maybe 'rdma' if one
> would be running ROCE. Shouldn't we error out on other transports like 'fc' or 'loop'?
> A2) This is only for TCP, and we do check that this option is only allowed for TCP by specifying it
> in the "allowed_opts" as follows (see file tcp.c):
> 
> static struct nvmf_transport_ops nvme_tcp_transport = {
>          .name           = "tcp",
>          .module         = THIS_MODULE,
>          .required_opts  = NVMF_OPT_TRADDR,
>          .allowed_opts   = NVMF_OPT_TRSVCID | NVMF_OPT_RECONNECT_DELAY |
>                            NVMF_OPT_HOST_TRADDR | NVMF_OPT_CTRL_LOSS_TMO |
>                            NVMF_OPT_HDR_DIGEST | NVMF_OPT_DATA_DIGEST |
>                            NVMF_OPT_NR_WRITE_QUEUES | NVMF_OPT_NR_POLL_QUEUES |
>                            NVMF_OPT_TOS | NVMF_OPT_HOST_TRIFACE,
>          .create_ctrl    = nvme_tcp_create_ctrl,
> };
>
Right, okay.

> Q3) Normally the options are just parts of the 'address' string; why didn't
> you use that approach here?
> A3) I don't understand what you mean?
> 
The 'address' string contains all transport arguments (without the 
transport type), concatenated by a ','.
So the 'host_iface' value should just be added to the address string 
itself, and not added as separate argument to the sysfs printf() call.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                Kernel Storage Architect
hare@suse.de                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer

_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme