From mboxrd@z Thu Jan 1 00:00:00 1970 From: Peter Ujfalusi Subject: Re: [PATCH] ASoC: omap-pcm: Allow only formats with 1, 2, and 4 byte physical size Date: Tue, 3 Mar 2015 14:45:28 +0200 Message-ID: <54F5ACE8.2050702@ti.com> References: <1425382133-17120-1-git-send-email-jsarha@ti.com> <54F59DE1.2010406@ti.com> <20150303122208.GA28158@bitmer.com> Mime-Version: 1.0 Content-Type: text/plain; charset="windows-1252" Content-Transfer-Encoding: quoted-printable Return-path: In-Reply-To: <20150303122208.GA28158@bitmer.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: alsa-devel-bounces@alsa-project.org Sender: alsa-devel-bounces@alsa-project.org To: Jarkko Nikula Cc: liam.r.girdwood@linux.intel.com, alsa-devel@alsa-project.org, linux-omap@vger.kernel.org, broonie@kernel.org, Jyri Sarha List-Id: linux-omap@vger.kernel.org On 03/03/2015 02:22 PM, Jarkko Nikula wrote: > On Tue, Mar 03, 2015 at 01:41:21PM +0200, Peter Ujfalusi wrote: >> On 03/03/2015 01:28 PM, Jyri Sarha wrote: >>> sDMA support only transfer elements with 1, 2, and 4 byte physical >>> size. Initialize the pcm driver accordingly. >> >> Acked-by: Peter Ujfalusi >> >>> >>> Signed-off-by: Jyri Sarha >>> --- >>> sound/soc/omap/omap-pcm.c | 21 ++++++++++++++++++++- >>> 1 file changed, 20 insertions(+), 1 deletion(-) >>> >>> diff --git a/sound/soc/omap/omap-pcm.c b/sound/soc/omap/omap-pcm.c >>> index f4b05bc..e49ee23 100644 >>> --- a/sound/soc/omap/omap-pcm.c >>> +++ b/sound/soc/omap/omap-pcm.c >>> @@ -39,7 +39,7 @@ >>> #define pcm_omap1510() 0 >>> #endif >>> = >>> -static const struct snd_pcm_hardware omap_pcm_hardware =3D { >>> +static struct snd_pcm_hardware omap_pcm_hardware =3D { >>> .info =3D SNDRV_PCM_INFO_MMAP | >>> SNDRV_PCM_INFO_MMAP_VALID | >>> SNDRV_PCM_INFO_INTERLEAVED | >>> @@ -53,6 +53,24 @@ static const struct snd_pcm_hardware omap_pcm_hardwa= re =3D { >>> .buffer_bytes_max =3D 128 * 1024, >>> }; >>> = >>> +/* sDMA supports only 1, 2, and 4 byte transfer elements. */ >>> +static void omap_pcm_limit_supported_formats(void) >>> +{ >>> + int i; >>> + >>> + for (i =3D 0; i < SNDRV_PCM_FORMAT_LAST; i++) { >>> + switch (snd_pcm_format_physical_width(i)) { >>> + case 8: >>> + case 16: >>> + case 32: >>> + omap_pcm_hardware.formats |=3D (1LL << i); >>> + break; >>> + default: >>> + break; >>> + } >>> + } >>> +} >>> + > = > I wonder why these are set in a loop instead of setting formats statically omap-pcm as such can work with any formats which has 1, 2 or 4 bytes layout= in memory (and not only S16_LE, S32_LE). sDMA can not handle 3byte words so we can not handle those with omap-pcm. In contrast eDMA for example can handle 1, 2, 3 and 4 w/o issue. > and why this is needed after commit 60e21d287344 ("ASoC: omap: Don't set > unused struct snd_pcm_hardware fields")? Hrm, interesting. We also have: 2d38df128321 (ASoC: generic-dmaengine-pcm: Prepare formats mask for valid physical sample sizes) Which does the same thing for the generic dmaengine pcm users. Actually the information from the omap_pcm_hardware is copied to runtime.hw via snd_soc_set_runtime_hwparams() call, so it is not ignored. We need this masking to be done here since the omap-pcm is still mostly cus= tom code and do not plug easily to the generic code (ie, not at all). -- = P=E9ter