Message ID | 20211206095920.40552-1-kory.maincent@bootlin.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | ASoC: soc-core: add the driver component name to the component struc | expand |
On Mon, Dec 06, 2021 at 10:59:20AM +0100, Kory Maincent wrote: > If a non i2c driver register two components the function will return the > same "device_name" for both components. This could cause unexpected issue, > in my case it is a debugfs error which tries to create two directory with > the same component name. Why is one device registering multiple components in the first place?
Hello Mark, On Mon, 6 Dec 2021 19:33:58 +0000 Mark Brown <broonie@kernel.org> wrote: > On Mon, Dec 06, 2021 at 10:59:20AM +0100, Kory Maincent wrote: > > > If a non i2c driver register two components the function will return the > > same "device_name" for both components. This could cause unexpected issue, > > in my case it is a debugfs error which tries to create two directory with > > the same component name. > > Why is one device registering multiple components in the first place? Because the sound components are more and more complex. Why they shouldn't? It seems to be already the case: sound/soc/codecs/cros_ec_codec.c sound/soc/fsl/fsl_easrc.c sound/soc/mediatek/mt*/mt*-afe-pcm.c sound/soc/sunxi/sun4i-codec.c sound/soc/soc-utils.c Regards, Köry
On Tue, Dec 07, 2021 at 09:47:32AM +0100, Köry Maincent wrote: > Mark Brown <broonie@kernel.org> wrote: > > Why is one device registering multiple components in the first place? > Because the sound components are more and more complex. Why they shouldn't? In what way are they more complex? > It seems to be already the case: > sound/soc/codecs/cros_ec_codec.c > sound/soc/fsl/fsl_easrc.c > sound/soc/mediatek/mt*/mt*-afe-pcm.c > sound/soc/sunxi/sun4i-codec.c > sound/soc/soc-utils.c Quite a few (I think all?) of these are quite old and and are the result of refactoring from pre-component code rather than modern drivers, it's likely there is no concrete reason for them to behave as they do.
Mark, On Tue, 7 Dec 2021 13:08:33 +0000 Mark Brown <broonie@kernel.org> wrote: > On Tue, Dec 07, 2021 at 09:47:32AM +0100, Köry Maincent wrote: > > Mark Brown <broonie@kernel.org> wrote: > > > > Why is one device registering multiple components in the first place? > > > Because the sound components are more and more complex. Why they shouldn't? > > > > In what way are they more complex? The sound hardware components add more and more features. > > > It seems to be already the case: > > sound/soc/codecs/cros_ec_codec.c > > sound/soc/fsl/fsl_easrc.c > > sound/soc/mediatek/mt*/mt*-afe-pcm.c > > sound/soc/sunxi/sun4i-codec.c > > sound/soc/soc-utils.c > > Quite a few (I think all?) of these are quite old and and are the result > of refactoring from pre-component code rather than modern drivers, it's > likely there is no concrete reason for them to behave as they do. I am a beginner in the kernel sound stack, alright then, the issue comes from the drivers. Thanks, Regards
diff --git a/sound/soc/soc-core.c b/sound/soc/soc-core.c index dcf6be4c4aaa..21ff77b231b8 100644 --- a/sound/soc/soc-core.c +++ b/sound/soc/soc-core.c @@ -2342,10 +2342,10 @@ EXPORT_SYMBOL_GPL(snd_soc_unregister_card); * Simplify DAI link configuration by removing ".-1" from device names * and sanitizing names. */ -static char *fmt_single_name(struct device *dev, int *id) +static char *fmt_single_name(struct device *dev, const char *snd_drv_name, int *id) { const char *devname = dev_name(dev); - char *found, *name; + char *found, *name, *tmp; unsigned int id1, id2; if (devname == NULL) @@ -2380,6 +2380,14 @@ static char *fmt_single_name(struct device *dev, int *id) *id = 0; } + if (snd_drv_name != NULL) { + /* Add driver component name if present */ + tmp = devm_kasprintf(dev, GFP_KERNEL, "%s.%s", snd_drv_name, name); + devm_kfree(dev, name); + name = devm_kstrdup(dev, tmp, GFP_KERNEL); + devm_kfree(dev, tmp); + } + return name; } @@ -2444,7 +2452,7 @@ struct snd_soc_dai *snd_soc_register_dai(struct snd_soc_component *component, */ if (legacy_dai_naming && (dai_drv->id == 0 || dai_drv->name == NULL)) { - dai->name = fmt_single_name(dev, &dai->id); + dai->name = fmt_single_name(dev, dai_drv->name, &dai->id); } else { dai->name = fmt_multiple_name(dev, dai_drv); if (dai_drv->id) @@ -2578,7 +2586,7 @@ int snd_soc_component_initialize(struct snd_soc_component *component, INIT_LIST_HEAD(&component->list); mutex_init(&component->io_mutex); - component->name = fmt_single_name(dev, &component->id); + component->name = fmt_single_name(dev, driver->name, &component->id); if (!component->name) { dev_err(dev, "ASoC: Failed to allocate name\n"); return -ENOMEM;
The function fmt_single_name is using to sanitize namings, and it return the name set to component->name. This function is returning either the "device_name" or the "device_name.driver_name" for i2c devices. There is no use of the component driver name. If a non i2c driver register two components the function will return the same "device_name" for both components. This could cause unexpected issue, in my case it is a debugfs error which tries to create two directory with the same component name. I have fixed it by prefixing the component name with the driver component name. Signed-off-by: Kory Maincent <kory.maincent@bootlin.com> --- sound/soc/soc-core.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-)