On Thu, Feb 26, 2026 at 08:58:51AM -0800, T.J. Mercier wrote:
> On Thu, Feb 26, 2026 at 2:38 AM Maxime Ripard <mripard@kernel.org> wrote:
> >
> > Hi David,
> >
> > On Thu, Feb 26, 2026 at 11:25:24AM +0100, David Hildenbrand (Arm) wrote:
> > > On 2/25/26 17:41, Maxime Ripard wrote:
> > > > The CMA dma-buf heap uses cma_alloc() and cma_release() to allocate and
> > > > free, respectively, its CMA buffers.
> > > >
> > > > However, these functions are not exported. Since we want to turn the CMA
> > > > heap into a module, let's export them both.
> > > >
> > > > Signed-off-by: Maxime Ripard <mripard@kernel.org>
> > > > ---
> > > >  mm/cma.c | 2 ++
> > > >  1 file changed, 2 insertions(+)
> > > >
> > > > diff --git a/mm/cma.c b/mm/cma.c
> > > > index 94b5da468a7d719e5144d33b06bcc7619c0fbcc9..be142b473f3bd41b9c7d8ba4397f018f6993d962 100644
> > > > --- a/mm/cma.c
> > > > +++ b/mm/cma.c
> > > > @@ -949,10 +949,11 @@ struct page *cma_alloc(struct cma *cma, unsigned long count,
> > > >     if (page)
> > > >             set_pages_refcounted(page, count);
> > > >
> > > >     return page;
> > > >  }
> > > > +EXPORT_SYMBOL_GPL(cma_alloc);
> > > >
> > > >  static struct cma_memrange *find_cma_memrange(struct cma *cma,
> > > >             const struct page *pages, unsigned long count)
> > > >  {
> > > >     struct cma_memrange *cmr = NULL;
> > > > @@ -1025,10 +1026,11 @@ bool cma_release(struct cma *cma, const struct page *pages,
> > > >
> > > >     __cma_release_frozen(cma, cmr, pages, count);
> > > >
> > > >     return true;
> > > >  }
> > > > +EXPORT_SYMBOL_GPL(cma_release);
> > > >
> > > >  bool cma_release_frozen(struct cma *cma, const struct page *pages,
> > > >             unsigned long count)
> > > >  {
> > > >     struct cma_memrange *cmr;
> > > >
> > >
> > > I'm wondering whether we want to restrict all these exports to the
> > > dma-buf module only using EXPORT_SYMBOL_FOR_MODULES().
> >
> > TIL about EXPORT_SYMBOL_FOR_MODULES, thanks.
> 
> Ohh, ditto.
> 
> > > Especially dma_contiguous_default_area() (patch #4), I am not sure
> > > whether we want arbitrary modules to mess with that.
> >
> > Yeah, I wasn't too fond about that one either. Alternatively, I guess we
> > could turn dev_get_cma_area into a non-inlined function and export that
> > instead?
> 
> I checked the history to see if dev_get_cma_area was converted to
> inline at some point for performance, but it has always been that way
> since 3.5. That'd be my only worry with un-inlining and exporting it.
> EXPORT_SYMBOL_FOR_MODULES sounds like a better way to me.

I just realised the new tegra heap is also going to use cma_alloc:
https://lore.kernel.org/dri-devel/20260122161009.3865888-7-thierry.reding@kernel.org/

I exported dev_get_cma_area() like we discussed, but kept EXPORT_SYMBOL
for now in the new version.

Maxime