Re: [REPOST PATCH 1/8] fence: dma-buf cross-device synchronization (v17)

[Greg KH <gregkh@linuxfoundation.org>]

On Thu, Jun 19, 2014 at 01:45:30PM -0400, Rob Clark wrote:
> On Thu, Jun 19, 2014 at 1:00 PM, Greg KH <gregkh@linuxfoundation.org> wrote:
> > On Thu, Jun 19, 2014 at 10:00:18AM -0400, Rob Clark wrote:
> >> On Wed, Jun 18, 2014 at 9:13 PM, Greg KH <gregkh@linuxfoundation.org> wrote:
> >> > On Wed, Jun 18, 2014 at 12:36:54PM +0200, Maarten Lankhorst wrote:
> >> >> +#define CREATE_TRACE_POINTS
> >> >> +#include <trace/events/fence.h>
> >> >> +
> >> >> +EXPORT_TRACEPOINT_SYMBOL(fence_annotate_wait_on);
> >> >> +EXPORT_TRACEPOINT_SYMBOL(fence_emit);
> >> >
> >> > Are you really willing to live with these as tracepoints for forever?
> >> > What is the use of them in debugging?  Was it just for debugging the
> >> > fence code, or for something else?
> >> >
> >> >> +/**
> >> >> + * fence_context_alloc - allocate an array of fence contexts
> >> >> + * @num:     [in]    amount of contexts to allocate
> >> >> + *
> >> >> + * This function will return the first index of the number of fences allocated.
> >> >> + * The fence context is used for setting fence->context to a unique number.
> >> >> + */
> >> >> +unsigned fence_context_alloc(unsigned num)
> >> >> +{
> >> >> +     BUG_ON(!num);
> >> >> +     return atomic_add_return(num, &fence_context_counter) - num;
> >> >> +}
> >> >> +EXPORT_SYMBOL(fence_context_alloc);
> >> >
> >> > EXPORT_SYMBOL_GPL()?  Same goes for all of the exports in here.
> >> > Traditionally all of the driver core exports have been with this
> >> > marking, any objection to making that change here as well?
> >>
> >> tbh, I prefer EXPORT_SYMBOL()..  well, I'd prefer even more if there
> >> wasn't even a need for EXPORT_SYMBOL_GPL(), but sadly it is a fact of
> >> life.  We already went through this debate once with dma-buf.  We
> >> aren't going to change $evil_vendor's mind about non-gpl modules.  The
> >> only result will be a more flugly convoluted solution (ie. use syncpt
> >> EXPORT_SYMBOL() on top of fence EXPORT_SYMBOL_GPL()) just as a
> >> workaround, with the result that no-one benefits.
> >
> > It has been proven that using _GPL() exports have caused companies to
> > release their code "properly" over the years, so as these really are
> > Linux-only apis, please change them to be marked this way, it helps
> > everyone out in the end.
> 
> Well, maybe that is the true in some cases.  But it certainly didn't
> work out that way for dma-buf.  And I think the end result is worse.
> 
> I don't really like coming down on the side of EXPORT_SYMBOL() instead
> of EXPORT_SYMBOL_GPL(), but if we do use EXPORT_SYMBOL_GPL() then the
> result will only be creative workarounds using the _GPL symbols
> indirectly by whatever is available via EXPORT_SYMBOL().  I don't
> really see how that will be better.

You are saying that you _know_ companies will violate our license, so
you should just "give up"?  And how do you know people aren't working on
preventing those "indirect" usages as well?  :)

Sorry, I'm not going to give up here, again, it has proven to work in
the past in changing the ways of _very_ large companies, why stop now?

thanks,

greg k-h