Re: [PATCH net-next 5/6] devlink: Reshuffle resource registration logic

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Fri, 19 Nov 2021 17:38:53 +0200 Leon Romanovsky wrote:
> On Thu, Nov 18, 2021 at 05:48:13PM -0800, Jakub Kicinski wrote:
> > On Thu, 18 Nov 2021 09:50:20 +0200 Leon Romanovsky wrote:  
> > > And it shouldn't. devlink_resource_find() will return valid resource only
> > > if there driver is completely bogus with races or incorrect allocations of
> > > resource_id.
> > > 
> > > devlink_*_register(..)
> > >  mutex_lock(&devlink->lock);
> > >  if (devlink_*_find(...)) {
> > >     mutex_unlock(&devlink->lock);
> > >     return ....;
> > >  }
> > >  .....
> > > 
> > > It is almost always wrong from locking and layering perspective the pattern above,
> > > as it is racy by definition if not protected by top layer.
> > > 
> > > There are exceptions from the rule above, but devlink is clearly not the
> > > one of such exceptions.  
> > 
> > Just drop the unnecessary "cleanup" patches and limit the amount 
> > of driver code we'll have to revert if your approach fails.  
> 
> My approach works, exactly like it works in other subsystems.
> https://lore.kernel.org/netdev/cover.1636390483.git.leonro@xxxxxxxxxx/

What "other subsystems"? I'm aware of the RFC version of these patches.

Breaking up the locks to to protect sub-objects only is fine for
protecting internal lists but now you can't guarantee that the object
exists when driver is called.

I'm sure you'll utter your unprovable "in real drivers.." but the fact
is my approach does not suffer from any such issues. Or depends on
drivers registering devlink last.

I can start passing a pointer to a devlink_port to split/unsplit
functions, which is a great improvement to the devlink driver API.

> We are waiting to see your proposal extended to support parallel devlink
> execution and to be applied to real drivers.
> https://lore.kernel.org/netdev/20211030231254.2477599-1-kuba@xxxxxxxxxx/

The conversion to xarray you have done is a great improvement, I don't
disagree with the way you convert to allow parallel calls either.

I already told you that real drivers can be converted rather easily,
even if it's not really necessary.

But I'm giving you time to make your proposal. If I spend time
polishing my patches I'll be even more eager to put this behind me.

> Anyway, you are maintainer, you want half work, you will get half work.

What do you mean half work? You have a record of breaking things 
in the area and changing directions. How is my request to limit
unnecessary "cleanups" affecting drivers until the work is finished
not perfectly reasonable?!?!

> > I spent enough time going back and forth with you.
> 
> Disagreements are hard for everyone, not only for you.



[Index of Archives]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Photo]     [Yosemite News]     [Yosemite Photos]     [Linux Kernel]     [Linux SCSI]     [XFree86]

  Powered by Linux