Re: [PATCH V1 2/3] ALSA: hda: add bus callback for address translation

[Takashi Iwai <tiwai@suse.de>]

On Mon, 01 Dec 2025 09:14:57 +0100,
Joakim  Zhang wrote:
> 
> 
> Hello Takashi,
> 
> > -----Original Message-----
> > From: Takashi Iwai <tiwai@suse.de>
> > Sent: Saturday, November 29, 2025 1:23 AM
> > To: Joakim Zhang <joakim.zhang@cixtech.com>
> > Cc: Takashi Iwai <tiwai@suse.de>; lgirdwood@gmail.com;
> > broonie@kernel.org; robh@kernel.org; krzk+dt@kernel.org;
> > conor+dt@kernel.org; perex@perex.cz; tiwai@suse.com; linux-
> > sound@vger.kernel.org; devicetree@vger.kernel.org; cix-kernel-upstream
> > <cix-kernel-upstream@cixtech.com>
> > Subject: Re: [PATCH V1 2/3] ALSA: hda: add bus callback for address
> > translation
> >
> > EXTERNAL EMAIL
> >
> > CAUTION: Suspicious Email from unusual domain.
> >
> > On Thu, 27 Nov 2025 10:44:26 +0100,
> > Joakim  Zhang wrote:
> > >
> > > [...]
> > > > >  include/sound/hdaudio.h     |  3 +++
> > > > >  sound/hda/core/controller.c | 25 +++++++++++++++++++------
> > > > >  sound/hda/core/stream.c     | 17 ++++++++++++++---
> > > > >  3 files changed, 36 insertions(+), 9 deletions(-)
> > > > >
> > > > > diff --git a/include/sound/hdaudio.h b/include/sound/hdaudio.h
> > > > > index 4e0c1d8af09f..61b41a014f4a 100644
> > > > > --- a/include/sound/hdaudio.h
> > > > > +++ b/include/sound/hdaudio.h
> > > > > @@ -293,6 +293,9 @@ struct hdac_bus {
> > > > >       const struct hdac_bus_ops *ops;
> > > > >       const struct hdac_ext_bus_ops *ext_ops;
> > > > >
> > > > > +     /* address translation from host to hdac */
> > > > > +     dma_addr_t (*addr_host_to_hdac)(struct hdac_bus *bus,
> > > > > + dma_addr_t addr);
> > > >
> > > > This should be rather added to hdac_bus_ops instead.
> > > >
> > > > Or, we can just add addr_offset field in hdac_bus instead of yet
> > > > another callback.  Then the change would be simpler.
> > > >
> > >
> > > Hello Takashi,
> > >
> > > Sorry for late updating since I am involving in other tasks recently.
> > >
> > > I tried to add addr_offset field in hdac_bus for the address
> > > translation, but met the issue, e.g.
> > > On our platform, if HOST address is 0xd0c09000, HDAC address is
> > 0xd0c09000 - 0x90000000 = 0x40c09000.
> > >
> > > If we use the pattern, HOST_addr + addr_offset = HDAC_addr For 64bit
> > > host, then addr_offset shound be 0x70000000,
> > > 0xd0c09000+0x70000000=0x1_40c09000, it will update both lower 32bit
> > > and upper 32bit into hda registers. However, hda controller in our
> > > audio subsystem can only support 32bit address, if hda  sending address
> > more than 32bit, it would not work.
> >
> > Hm, I don't understand.  It we define something like
> >
> >         dma_addr_t addr_offset;
> >
> > in hdac_bus, and the driver setting
> >
> >         bus->addr_offset = -0x90000000;
> >
> > then a calculation
> >
> >         dma_addr_t addr;
> >
> >         addr = bus->rb.addr + bus->addr_offset;
> >
> > will lead to the very same result as your callback, no?
> > Or am I missing something?
> 
> typedef u64 dma_addr_t;
> 
> 1)
> dma_addr_t a = 0xd0c09000;
> dma_addr_t b = -0x90000000;

You have to pass with ULL suffix as a 64bit value instead:
  dma_addr_t b = -0x90000000ULL;


thanks,

Takashi