Re: [PATCH 2/2] dmabuf/tracing: Add dma-buf trace events

[Suren Baghdasaryan]

On Mon, Aug 3, 2020 at 6:09 PM Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
>
> On Mon, Aug 03, 2020 at 11:28:31PM +0100, Al Viro wrote:
>
> > IOW, what the hell is that horror for?  You do realize, for example, that there's
> > such thing as dup(), right?  And dup2() as well.  And while we are at it, how
> > do you keep track of removals, considering the fact that you can stick a file
> > reference into SCM_RIGHTS datagram sent to yourself, close descriptors and an hour
> > later pick that datagram, suddenly getting descriptor back?
> >
> > Besides, "I have no descriptors left" != "I can't be currently sitting in the middle
> > of syscall on that sucker"; close() does *NOT* terminate ongoing operations.

Thanks for your feedback, Al. I see your points and sorry for not
realizing these shortcomings.

> >
> > You are looking at the drastically wrong abstraction level.  Please, describe what
> > it is that you are trying to achieve.
>
> _IF_ it's "who keeps a particularly long-lived sucker pinned", I would suggest
> fuser(1) run when you detect that kind of long-lived dmabuf.  With events generated
> by their constructors and destructors, and detection of longevity done based on
> that.

That is the intention here. IIUC fuser(1) would require root access to
collect this information from a process other than the caller. Ideally
what we would like to have is a non-root process with specific
capabilities (in our case a process that can access BPF maps) to be
able to obtain the information on dma-buf users.
However, it might make more sense to track dma-buf usage from
dma_buf_getfile, dma_buf_get and dma_buf_put since these calls are the
ones that affect file refcount. Will dig some more into this.
Thanks for your time and sorry for not thinking it through beforehand.

>
> But that's only a semi-blind guess at the things you are trying to achieve; please,
> describe what it really is.