[PATCH 1/2] iouring: one capable call per iouring instance

[Ming Lei]

On Tue, Dec 05, 2023 at 08:45:10AM -0700, Keith Busch wrote:
> On Tue, Dec 05, 2023 at 01:25:44PM +0800, Ming Lei wrote:
> > On Mon, Dec 04, 2023 at 09:31:21PM -0700, Keith Busch wrote:
> > > Good question. The "capable" check had always been first so even with
> > > the relaxed permissions, it was still paying the price. I have changed
> > > that order in commit staged here (not yet upstream):
> > > 
> > >   http://git.infradead.org/nvme.git/commitdiff/7be866b1cf0bf1dfa74480fe8097daeceda68622
> > 
> > With this change, I guess you shouldn't see the following big gap, right?
> 
> Correct.
>  
> > > Before: 970k IOPs
> > > After: 1750k IOPs
>  
> > > Note that only prevents the costly capable() check if the inexpensive
> > > checks could make a determination. That's still not solving the problem
> > > long term since we aim for forward compatibility where we have no idea
> > > which opcodes, admin identifications, or vendor specifics could be
> > > deemed "safe" for non-root users in the future, so those conditions
> > > would always fall back to the more expensive check that this patch was
> > > trying to mitigate for admin processes.
> > 
> > Not sure I get the idea, it is related with nvme's permission model for
> > user pt command, and:
> > 
> > 1) it should be always checked in entry of nvme user pt command
> > 
> > 2) only the following two types of commands require ADMIN, per commit
> > 855b7717f44b ("nvme: fine-granular CAP_SYS_ADMIN for nvme io commands")
> > 
> >     - any admin-cmd is not allowed
> >     - vendor-specific and fabric commmand are not allowed
> > 
> > Can you provide more details why the expensive check can't be avoided for
> > fast read/write user IO commands?
> 
> It's not necessarily about the read/write passthrough commands. It's for
> commands we don't know about today. Do we want to revisit this problem
> every time spec provides another operation? Are vendor unique solutions
> not allowed to get high IOPs access?

Except for read/write, what other commands are performance sensitive?

> 
> Secondly, some people have rediscovered you can abuse this interface to
> corrupt kernel memory, so there are considerations to restricting this

Just wondering why ADMIN won't corrupt kernel memory, and only normal
user can, looks it is kernel bug instead of permission related issue.

> to CAP_SYS_ADMIN anyway, so there's no cheap check available today if we
> have to go that route.

If capable(CAP_SYS_ADMIN) is really slow, I am wondering why not
optimize it in task_struct?


Thanks, 
Ming