Re: [PATCH v3 2/2] misc: sram: Add DMA-BUF Heap exporting of SRAM areas

[Greg Kroah-Hartman]

On Fri, Apr 19, 2024 at 06:57:47PM +0200, Christian Gmeiner wrote:
> Am Di., 9. Apr. 2024 um 14:14 Uhr schrieb Greg Kroah-Hartman
> <gregkh@xxxxxxxxxxxxxxxxxxx>:
> >
> > On Tue, Apr 09, 2024 at 02:06:05PM +0200, Pascal FONTAIN wrote:
> > > From: Andrew Davis <afd@xxxxxx>
> > >
> > > This new export type exposes to userspace the SRAM area as a DMA-BUF
> > > Heap,
> > > this allows for allocations of DMA-BUFs that can be consumed by various
> > > DMA-BUF supporting devices.
> > >
> > > Signed-off-by: Andrew Davis <afd@xxxxxx>
> > > Tested-by: Pascal Fontain <pascal.fontain@xxxxxxxxxxxxx>
> >
> > When sending on a patch from someone else, you too must sign off on it
> > as per our documenation.  Please read it and understand what you are
> > agreeing to when you do that for a new version please.
> >
> > > ---
> > >  drivers/misc/Kconfig         |   7 +
> > >  drivers/misc/Makefile        |   1 +
> > >  drivers/misc/sram-dma-heap.c | 246 +++++++++++++++++++++++++++++++++++
> > >  drivers/misc/sram.c          |   6 +
> > >  drivers/misc/sram.h          |  16 +++
> > >  5 files changed, 276 insertions(+)
> > >  create mode 100644 drivers/misc/sram-dma-heap.c
> > >
> > > diff --git a/drivers/misc/Kconfig b/drivers/misc/Kconfig
> > > index 9e4ad4d61f06..e6674e913168 100644
> > > --- a/drivers/misc/Kconfig
> > > +++ b/drivers/misc/Kconfig
> > > @@ -448,6 +448,13 @@ config SRAM
> > >  config SRAM_EXEC
> > >       bool
> > >
> > > +config SRAM_DMA_HEAP
> > > +     bool "Export on-chip SRAM pools using DMA-Heaps"
> > > +     depends on DMABUF_HEAPS && SRAM
> > > +     help
> > > +       This driver allows the export of on-chip SRAM marked as both pool
> > > +       and exportable to userspace using the DMA-Heaps interface.
> >
> > What will use this in userspace?
> >
> 
> I could imagine a way it might be used.

This implies it is not needed because no one actually has actually used
it for anything, so why make this change?

> Imagine a SoC like TI's AM64x series, where some cores (A53) run Linux
> and others (R5F) are managed by remoteproc to
> execute a RTOS. When it comes to efficiently exchanging large data
> sets, the conventional method involves using rpmsg,
> which has limitations due to message size and could potentially slow
> down data transfer. However, an alternative
> approach could be to allocate a sizable chunk of SRAM memory in
> userspace. By utilizing memcpy() to copy data into
> this memory, followed by a single rpmsg signal to notify the RTOS that
> the data is ready, we can leverage the faster access
> speed of SRAM compared to DDR from the remoteproc.

Why is virtio not used instead?

thanks,

greg k-h