[alsa-devel] [PATCH 2.6.37 1/2] ASoC: Fix dapm_seq_compare() for multi-component
Ensure that we keep all widget powerups in DAPM sequence by making the CODEC the last thing we compare on rather than the first thing. Also fix the fact that we're currently comparing the widget pointers rather than the CODEC pointers when we do the substraction so we won't get stable results.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com --- sound/soc/soc-dapm.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/sound/soc/soc-dapm.c b/sound/soc/soc-dapm.c index 8352430..bc2ec06 100644 --- a/sound/soc/soc-dapm.c +++ b/sound/soc/soc-dapm.c @@ -706,12 +706,12 @@ static int dapm_seq_compare(struct snd_soc_dapm_widget *a, struct snd_soc_dapm_widget *b, int sort[]) { - if (a->codec != b->codec) - return (unsigned long)a - (unsigned long)b; if (sort[a->id] != sort[b->id]) return sort[a->id] - sort[b->id]; if (a->reg != b->reg) return a->reg - b->reg; + if (a->codec != b->codec) + return (unsigned long)a->codec - (unsigned long)b->codec;
return 0; }
DAPM widgets may be associated with non-CODEC devices so compare based on the DAPM context rather than the CODEC pointer.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com --- sound/soc/soc-dapm.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/sound/soc/soc-dapm.c b/sound/soc/soc-dapm.c index bc2ec06..5ee93a5 100644 --- a/sound/soc/soc-dapm.c +++ b/sound/soc/soc-dapm.c @@ -710,8 +710,8 @@ static int dapm_seq_compare(struct snd_soc_dapm_widget *a, return sort[a->id] - sort[b->id]; if (a->reg != b->reg) return a->reg - b->reg; - if (a->codec != b->codec) - return (unsigned long)a->codec - (unsigned long)b->codec; + if (a->dapm != b->dapm) + return (unsigned long)a->dapm - (unsigned long)b->dapm;
return 0; }
On Fri, 12 Nov 2010 16:01:11 +0000 Mark Brown broonie@opensource.wolfsonmicro.com wrote:
DAPM widgets may be associated with non-CODEC devices so compare based on the DAPM context rather than the CODEC pointer.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com
sound/soc/soc-dapm.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/sound/soc/soc-dapm.c b/sound/soc/soc-dapm.c index bc2ec06..5ee93a5 100644 --- a/sound/soc/soc-dapm.c +++ b/sound/soc/soc-dapm.c @@ -710,8 +710,8 @@ static int dapm_seq_compare(struct snd_soc_dapm_widget *a, return sort[a->id] - sort[b->id]; if (a->reg != b->reg) return a->reg - b->reg;
- if (a->codec != b->codec)
return (unsigned long)a->codec - (unsigned long)b->codec;
- if (a->dapm != b->dapm)
return (unsigned long)a->dapm - (unsigned long)b->dapm;
Definitely, this conversion was missing from DAPM decoupling patch.
Acked-by: Jarkko Nikula jhnikula@gmail.com
On Fri, 2010-11-12 at 16:01 +0000, Mark Brown wrote:
DAPM widgets may be associated with non-CODEC devices so compare based on the DAPM context rather than the CODEC pointer.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com
Both
Acked-by: Liam Girdwood lrg@slimlogic.co.uk
On Fri, 12 Nov 2010 16:01:10 +0000 Mark Brown broonie@opensource.wolfsonmicro.com wrote:
Ensure that we keep all widget powerups in DAPM sequence by making the CODEC the last thing we compare on rather than the first thing. Also fix the fact that we're currently comparing the widget pointers rather than the CODEC pointers when we do the substraction so we won't get stable results.
Signed-off-by: Mark Brown broonie@opensource.wolfsonmicro.com
sound/soc/soc-dapm.c | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/sound/soc/soc-dapm.c b/sound/soc/soc-dapm.c index 8352430..bc2ec06 100644 --- a/sound/soc/soc-dapm.c +++ b/sound/soc/soc-dapm.c @@ -706,12 +706,12 @@ static int dapm_seq_compare(struct snd_soc_dapm_widget *a, struct snd_soc_dapm_widget *b, int sort[]) {
- if (a->codec != b->codec)
return (unsigned long)a - (unsigned long)b;
- if (a->codec != b->codec)
return (unsigned long)a->codec - (unsigned long)b->codec;
This sounds feasible change. What I was thinking are there any benefit which one, the register or codec is compared first but I don't think there's any practical difference.
A1, B1, A2, B2 (now) or A1, A2, B1, B2 (codec comparison before register)
Acked-by: Jarkko Nikula jhnikula@gmail.com
On Fri, Nov 12, 2010 at 09:06:37PM +0200, Jarkko Nikula wrote:
This sounds feasible change. What I was thinking are there any benefit which one, the register or codec is compared first but I don't think there's any practical difference.
A1, B1, A2, B2 (now) or A1, A2, B1, B2 (codec comparison before register)
If you've got two identical CODECs it will *probably* sound smoother to alternate between the two rather than enabling one then the other - there will be less time difference between any audible issues in each path. Probably a very small chance of that having an effect but it seemed more likely to do something useful than any other scenario.
participants (3)
-
Jarkko Nikula -
Liam Girdwood -
Mark Brown