netdev - Re: [PATCH 07/22] RDMA/irdma: Register an auxiliary driver and implement private channel OPs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20210127024423.GL4147@nvidia.com>
Date:   Tue, 26 Jan 2021 22:44:23 -0400
From:   Jason Gunthorpe <jgg@...dia.com>
To:     "Saleem, Shiraz" <shiraz.saleem@...el.com>
CC:     "Keller, Jacob E" <jacob.e.keller@...el.com>,
        Leon Romanovsky <leon@...nel.org>,
        "dledford@...hat.com" <dledford@...hat.com>,
        "kuba@...nel.org" <kuba@...nel.org>,
        "davem@...emloft.net" <davem@...emloft.net>,
        "linux-rdma@...r.kernel.org" <linux-rdma@...r.kernel.org>,
        "gregkh@...uxfoundation.org" <gregkh@...uxfoundation.org>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "Ertman, David M" <david.m.ertman@...el.com>,
        "Nguyen, Anthony L" <anthony.l.nguyen@...el.com>,
        "Ismail, Mustafa" <mustafa.ismail@...el.com>,
        "jiri@...dia.com" <jiri@...dia.com>,
        "Samudrala, Sridhar" <sridhar.samudrala@...el.com>,
        "Williams, Dan J" <dan.j.williams@...el.com>
Subject: Re: [PATCH 07/22] RDMA/irdma: Register an auxiliary driver and
 implement private channel OPs

On Wed, Jan 27, 2021 at 12:42:09AM +0000, Saleem, Shiraz wrote:

> > It does, the PCI driver is not supposed to spawn any aux devices for RDMA at all
> > if RDMA is disabled.
> > 
> > For an iWarp driver I would consider ENABLE_ROCE to really be a general
> > ENABLE_RDMA.
> 
> Well the driver supports iWARP and RoCE for E810 device.
> Are you saying that this generic enable_roce devlink param really
> is an enable 'rdma' traffic or not param?

I've thought of it that way, that is what it was created for at least.

Overloading it to be a iwarp not roce switch feels wrong
 
> > Are you sure you need to implement this?
> 
> What we are after is some mechanism for user to switch the protocols iWARP vs RoCE
> [default the device comes up as an iWARP dev]. The protocol info is really needed early-on
> in the RDMA driver.probe(). i.e. when the rdma admin queue is created.

This needs to be a pci devlink at least, some kind of mode switch
seems appropriate
 
> The same goes with the other param resource_limits_selector. It's a
> profile selector that a user can chose to different # of max QP,
> CQs, MRs etc.

And it must be done at init time? Also seems like pci devlink

Generally speaking anything that requires the rdma driver to be
reloaded should remove/restore the aux device.

Mode switch from roce to/from iwarp should create aux devices of
different names which naturally triggers the right kind of sequences
in the driver core

Jason