On Thu, Jul 25, 2019 at 09:29:40PM -0700, Sagi Grimberg wrote:
> 
> > > > > > > > NVMe-OF is configured using configfs. The target is specified 
> > > > > > > > by the
> > > > > > > > user writing a path to a configfs attribute. This is the way it 
> > > > > > > > works
> > > > > > > > today but with blkdev_get_by_path()[1]. For the passthru code, 
> > > > > > > > we need
> > > > > > > > to get a nvme_ctrl instead of a block_device, but the principal 
> > > > > > > > is the same.
> > > > > > > 
> > > > > > > Why isn't a fd being passed in there instead of a random string?
> > > > > > 
> > > > > > I suppose we could echo a string of the file descriptor number 
> > > > > > there,
> > > > > > and look up the fd in the process' file descriptor table ...
> > > > > 
> > > > > Assuming that there is a open handle somewhere out there...
> > > 
> > > Yes, that would be a step backwards from an interface. The user would
> > > then need a special process to open the fd and pass it through configfs.
> > > They couldn't just do it with basic bash commands.
> > 
> > First of all, they can, but... WTF not have filp_open() done right there?
> > Yes, by name.  With permission checks done.  And pick your object from the
> > sodding struct file you'll get.
> > 
> > What's the problem?  Why do you need cdev lookups, etc., when you are
> > dealing with files under your full control?  Just open them and use
> > ->private_data or whatever you set in ->open() to access the damn thing.
> > All there is to it...
> Oh this is so much simpler. There is really no point in using anything
> else. Just need to remember to compare f->f_op to what we expect to make
> sure that it is indeed the same device class.

That is good, that solves the "/dev/random/" issue I was talking about
earlier as well.

Odds are you all can do the same for the blockdev interface as well.

thanks,

greg k-h

Reply via email to