[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20241211160055.GM1888283@ziepe.ca>
Date: Wed, 11 Dec 2024 12:00:55 -0400
From: Jason Gunthorpe <jgg@...pe.ca>
To: Jakub Kicinski <kuba@...nel.org>
Cc: Bernard Metzler <bmt@...ich.ibm.com>, linux-rdma@...r.kernel.org,
leon@...nel.org, linux-kernel@...r.kernel.org,
netdev@...r.kernel.org, syzkaller-bugs@...glegroups.com,
zyjzyj2000@...il.com,
syzbot+4b87489410b4efd181bf@...kaller.appspotmail.com
Subject: Re: [PATCH] RDMA/siw: Remove direct link to net_device
On Tue, Dec 10, 2024 at 05:52:37PM -0800, Jakub Kicinski wrote:
> On Tue, 10 Dec 2024 10:56:27 -0400 Jason Gunthorpe wrote:
> > > struct siw_device {
> > > struct ib_device base_dev;
> > > - struct net_device *netdev;
> > > struct siw_dev_cap attrs;
> > >
> > > u32 vendor_part_id;
> > > + struct {
> > > + int ifindex;
> >
> > ifindex is only stable so long as you are holding a reference on the
> > netdev..
>
> Does not compute. Can you elaborate what you mean, Jason?
I mean you can't replace a netdev pointer with an ifindex, you can't
reliably get back to the same netdev from ifindex alone.
Jason
Powered by blists - more mailing lists