[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20250624185825.GA213144@ziepe.ca>
Date: Tue, 24 Jun 2025 15:58:25 -0300
From: Jason Gunthorpe <jgg@...pe.ca>
To: Alex Mastro <amastro@...com>
Cc: Alex Williamson <alex.williamson@...hat.com>, peterx@...hat.com,
kbusch@...nel.org, kvm@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] vfio/pci: print vfio-device name to fdinfo
On Tue, Jun 24, 2025 at 11:53:25AM -0700, Alex Mastro wrote:
> On Tue, 24 Jun 2025 15:16:16 -0300 Jason Gunthorpe <jgg@...pe.ca> wrote:
> > There will be a symlink under /sys/class/vfio-xx/XX pointing to the
> > <pci sysfs path>/vfio-dev/X directory
> >
> > And another symlink under /sys/dev/char/XX:XX doing the same.
>
> Got it, thanks. The issue does seem solved for the vfio cdev case, then.
I wonder if we could arrange things so that if the cdev path is turned
on then the device FD created by the ioctl would report the same
/dev/vfio/X information? I've never looked at how this works, but
maybe it is easy?
Even if userspace is not using the cdev it does exist and is still
logically affiliated with the ioctl FD.
Then we'd have a nice consistent user experience.
Jason
Powered by blists - more mailing lists