On Mon, Apr 01, 2019 at 09:25:44PM +0300, Leon Romanovsky wrote: > On Mon, Apr 01, 2019 at 02:47:42PM -0300, Jason Gunthorpe wrote: > > On Mon, Apr 01, 2019 at 05:18:16PM +0000, Parav Pandit wrote: > > > > > > > > > > From: Gal Pressman <galpress@xxxxxxxxxx> > > > > Sent: Thursday, March 28, 2019 7:39 AM > > > > To: Jason Gunthorpe <jgg@xxxxxxxx>; Doug Ledford <dledford@xxxxxxxxxx> > > > > Cc: Yossi Leybovich <sleybo@xxxxxxxxxx>; Alexander Matushevsky > > > > <matua@xxxxxxxxxx>; Leah Shalev <shalevl@xxxxxxxxxx>; Dave Goodell > > > > <goodell@xxxxxxxxxx>; Brian Barrett <bbarrett@xxxxxxxxxx>; linux- > > > > rdma@xxxxxxxxxxxxxxx; Sean Hefty <sean.hefty@xxxxxxxxx>; Dennis > > > > Dalessandro <dennis.dalessandro@xxxxxxxxx>; Leon Romanovsky > > > > <leon@xxxxxxxxxx>; Christoph Hellwig <hch@xxxxxxxxxxxxx>; Parav Pandit > > > > <parav@xxxxxxxxxxxx>; Sagi Grimberg <sagi@xxxxxxxxxxx>; Steve Wise > > > > <larrystevenwise@xxxxxxxxx>; Shiraz Saleem <shiraz.saleem@xxxxxxxxx>; > > > > Gal Pressman <galpress@xxxxxxxxxx>; Jason Baron <jbaron@xxxxxxxxxx> > > > > Subject: [PATCH rdma-next v4 01/12] RDMA/core: Introduce RDMA > > > > subsystem ibdev_* print functions > > > > > > > > Similarly to dev/netdev/etc printk helpers, add standard printk helpers for > > > > the RDMA subsystem. > > > > > > > > Example output: > > > > efa 0000:00:06.0 efa_0: Hello World! > > > > efa_0: Hello World! (no parent device set) (NULL ib_device): Hello World! > > > > (ibdev is NULL) > > > > > > > > > We are working on mdev devices whose device name in kernel defined > > > as 36 letters.. (standard uuid based) which is too long. > > > > Well, then you have a problem in netdev, so whatever is solved in > > netdev we should solve it here. Otherwise this should continue to copy > > netdev. > > > > I continue to think naming devices with random GUIDs is a really dumb > > thing to do. > > You are already doing it but in small scale by assigning some random > indexes to ib_device name, e.g. in efa_X and efa_X+1, X and X+1 are > GUIDs which mean nothing. They are not a GUID because they are not 'global'. The GUID's this mdev thing wants to use are actualy GUIDs and have enough characters to make them properly global if randomly sourced - which nothing actually needs or wants.. Jason
