[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20210325183655.GK2356281@nvidia.com>
Date: Thu, 25 Mar 2021 15:36:55 -0300
From: Jason Gunthorpe <jgg@...dia.com>
To: Keith Busch <kbusch@...nel.org>
Cc: Bjorn Helgaas <helgaas@...nel.org>,
Alexander Duyck <alexander.duyck@...il.com>,
Leon Romanovsky <leon@...nel.org>,
Bjorn Helgaas <bhelgaas@...gle.com>,
Saeed Mahameed <saeedm@...dia.com>,
Leon Romanovsky <leonro@...dia.com>,
Jakub Kicinski <kuba@...nel.org>,
linux-pci <linux-pci@...r.kernel.org>,
linux-rdma@...r.kernel.org, Netdev <netdev@...r.kernel.org>,
Don Dutile <ddutile@...hat.com>,
Alex Williamson <alex.williamson@...hat.com>,
"David S . Miller" <davem@...emloft.net>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Subject: Re: [PATCH mlx5-next v7 0/4] Dynamically assign MSI-X vectors count
On Fri, Mar 26, 2021 at 03:31:57AM +0900, Keith Busch wrote:
> The NVMe IO queue resources are assignable just like the MSIx vectors.
> But they're not always assigned 1:1. For example:
But this is all driver configuration, the driver might be running in
some VM.
It seems saner to have two kernel interfaces, one used in the
hypervisor on the PF to setup the MSI vector count
And a second interface that becomes available after the driver is
bound to configure the driver, where-ever/how-ever that driver may be
running.
I don't know how you could combine them in one step except for the
simple case of the driver running in the hypervisor?
Jason
Powered by blists - more mailing lists