Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1725368 > unrolled thread

[PATCH 00/11] mux/typec: Add USB / TypeC mux drivers and hook them up on some x86 systems

Started byHans de Goede <hdegoede@redhat.com>
First post2017-09-01 23:50 +0200
Last post2017-09-05 13:00 +0200
Articles 4 on this page of 24 — 5 participants

Back to article view | Back to linux.kernel

Contents

  [PATCH 00/11] mux/typec: Add USB / TypeC mux drivers and hook them up on some x86 systems Hans de Goede <hdegoede@redhat.com> - 2017-09-01 23:50 +0200
    [PATCH 03/11] mux: consumer.h: Add MUX_USB_* state constant defines Hans de Goede <hdegoede@redhat.com> - 2017-09-01 23:50 +0200
      Re: [PATCH 03/11] mux: consumer.h: Add MUX_USB_* state constant defines Andy Shevchenko <andy.shevchenko@gmail.com> - 2017-09-02 12:20 +0200
        Re: [PATCH 03/11] mux: consumer.h: Add MUX_USB_* state constant  defines Hans de Goede <hdegoede@redhat.com> - 2017-09-02 14:10 +0200
      Re: [PATCH 03/11] mux: consumer.h: Add MUX_USB_* state constant  defines Guenter Roeck <linux@roeck-us.net> - 2017-09-02 17:10 +0200
        Re: [PATCH 03/11] mux: consumer.h: Add MUX_USB_* state constant  defines Hans de Goede <hdegoede@redhat.com> - 2017-09-02 18:00 +0200
          Re: [PATCH 03/11] mux: consumer.h: Add MUX_USB_* state constant  defines Guenter Roeck <linux@roeck-us.net> - 2017-09-02 21:10 +0200
            Re: [PATCH 03/11] mux: consumer.h: Add MUX_USB_* state constant  defines Hans de Goede <hdegoede@redhat.com> - 2017-09-02 21:50 +0200
    [PATCH 08/11] staging: typec: tcpm: Set mux to device mode when configured as such Hans de Goede <hdegoede@redhat.com> - 2017-09-01 23:50 +0200
    [PATCH 07/11] extcon: intel-int3496: Add support for controlling the USB-role mux Hans de Goede <hdegoede@redhat.com> - 2017-09-01 23:50 +0200
      Re: [PATCH 07/11] extcon: intel-int3496: Add support for controlling  the USB-role mux Andy Shevchenko <andy.shevchenko@gmail.com> - 2017-09-02 12:40 +0200
        Re: [PATCH 07/11] extcon: intel-int3496: Add support for controlling  the USB-role mux Hans de Goede <hdegoede@redhat.com> - 2017-09-04 16:20 +0200
    [PATCH 10/11] staging: typec: fusb302: Hook up mux support using tcpc_gen_mux support Hans de Goede <hdegoede@redhat.com> - 2017-09-01 23:50 +0200
    [PATCH 01/11] mux: core: Add of_mux_control_get helper function Hans de Goede <hdegoede@redhat.com> - 2017-09-01 23:50 +0200
    [PATCH 09/11] staging: typec: Add Generic TCPC mux driver using the mux subsys Hans de Goede <hdegoede@redhat.com> - 2017-09-02 00:00 +0200
    [PATCH 04/11] usb: xhci: Add Intel cherrytrail extended cap / otg phy mux handling Hans de Goede <hdegoede@redhat.com> - 2017-09-02 00:00 +0200
      Re: [PATCH 04/11] usb: xhci: Add Intel cherrytrail extended cap /  otg phy mux handling Heikki Krogerus <heikki.krogerus@linux.intel.com> - 2017-09-04 09:40 +0200
        Re: [PATCH 04/11] usb: xhci: Add Intel cherrytrail extended cap / otg  phy mux handling Hans de Goede <hdegoede@redhat.com> - 2017-09-05 12:10 +0200
    [PATCH 06/11] mux: Add Pericom PI3USB30532 Type-C mux driver Hans de Goede <hdegoede@redhat.com> - 2017-09-02 00:00 +0200
    [PATCH 02/11] mux: core: Add support for getting a mux controller on a non DT platform Hans de Goede <hdegoede@redhat.com> - 2017-09-02 00:00 +0200
      Re: [PATCH 02/11] mux: core: Add support for getting a mux controller  on a non DT platform sathya <sathyaosid@gmail.com> - 2017-09-02 21:20 +0200
        Re: [PATCH 02/11] mux: core: Add support for getting a mux controller  on a non DT platform Hans de Goede <hdegoede@redhat.com> - 2017-09-04 16:30 +0200
      Re: [PATCH 02/11] mux: core: Add support for getting a mux controller  on a non DT platform Hans de Goede <hdegoede@redhat.com> - 2017-09-05 13:00 +0200
    Re: [PATCH 00/11] mux/typec: Add USB / TypeC mux drivers and hook  them up on some x86 systems Hans de Goede <hdegoede@redhat.com> - 2017-09-05 13:00 +0200

Page 2 of 2 — ← Prev page 1 [2]

#1725553 — Re: [PATCH 02/11] mux: core: Add support for getting a mux controller on a non DT platform

Fromsathya <sathyaosid@gmail.com>
Date2017-09-02 21:20 +0200
SubjectRe: [PATCH 02/11] mux: core: Add support for getting a mux controller on a non DT platform
Message-ID<ulm7w-6St-11@gated-at.bofh.it>
In reply to#1725383
Hi,


On 09/01/2017 02:48 PM, Hans de Goede wrote:
> On non DT platforms we cannot get the mux_chip by pnode. Other subsystems
> (regulator, clock, pwm) have the same problem and solve this by allowing
> platform / board-setup code to add entries to a lookup table and then use
> this table to look things up.
>
> This commit adds support for getting a mux controller on a non DT platform
> following this pattern. It is based on a simplified version of the pwm
> subsys lookup code, the dev_id and mux_name parts of a lookup table entry
> are mandatory in the mux-core implementation.
>
> Signed-off-by: Hans de Goede <hdegoede@redhat.com>
> ---
>   drivers/mux/core.c           | 96 +++++++++++++++++++++++++++++++++++++++++++-
>   include/linux/mux/consumer.h | 11 +++++
>   2 files changed, 106 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/mux/core.c b/drivers/mux/core.c
> index 6142493c327b..8864cc745506 100644
> --- a/drivers/mux/core.c
> +++ b/drivers/mux/core.c
> @@ -24,6 +24,9 @@
>   #include <linux/of_platform.h>
>   #include <linux/slab.h>
>   
> +static DEFINE_MUTEX(mux_lookup_lock);
> +static LIST_HEAD(mux_lookup_list);
> +
>   /*
>    * The idle-as-is "state" is not an actual state that may be selected, it
>    * only implies that the state should not be changed. So, use that state
> @@ -408,6 +411,23 @@ int mux_control_deselect(struct mux_control *mux)
>   }
>   EXPORT_SYMBOL_GPL(mux_control_deselect);
>   
> +static int parent_name_match(struct device *dev, const void *data)
> +{
> +	const char *parent_name = dev_name(dev->parent);
Device name usually contains id section (devname.id). Did you take this 
into consideration ?
> +	const char *name = data;
> +
> +	return strcmp(parent_name, name) == 0;
> +}
> +
> +static struct mux_chip *mux_chip_get_by_name(const char *name)
> +{
> +	struct device *dev;
> +
> +	dev = class_find_device(&mux_class, NULL, name, parent_name_match);
> +
> +	return dev ? to_mux_chip(dev) : NULL;
> +}
> +
>   static int of_dev_node_match(struct device *dev, const void *data)
>   {
>   	return dev->of_node == data;
> @@ -479,6 +499,42 @@ static struct mux_control *of_mux_control_get(struct device *dev,
>   }
>   
>   /**
> + * mux_add_table() - register PWM device consumers
PWM -> MUX
> + * @table: array of consumers to register
> + * @num: number of consumers in table
> + */
> +void mux_add_table(struct mux_lookup *table, size_t num)
> +{
> +	mutex_lock(&mux_lookup_lock);
> +
> +	while (num--) {
> +		list_add_tail(&table->list, &mux_lookup_list);
> +		table++;
> +	}
> +
> +	mutex_unlock(&mux_lookup_lock);
> +}
> +EXPORT_SYMBOL_GPL(mux_add_table);
> +
> +/**
> + * mux_remove_table() - unregister PWM device consumers
> + * @table: array of consumers to unregister
> + * @num: number of consumers in table
> + */
> +void mux_remove_table(struct mux_lookup *table, size_t num)
> +{
> +	mutex_lock(&mux_lookup_lock);
> +
> +	while (num--) {
> +		list_del(&table->list);
> +		table++;
> +	}
> +
> +	mutex_unlock(&mux_lookup_lock);
> +}
> +EXPORT_SYMBOL_GPL(mux_remove_table);
> +
> +/**
>    * mux_control_get() - Get the mux-control for a device.
>    * @dev: The device that needs a mux-control.
>    * @mux_name: The name identifying the mux-control.
> @@ -487,11 +543,49 @@ static struct mux_control *of_mux_control_get(struct device *dev,
>    */
>   struct mux_control *mux_control_get(struct device *dev, const char *mux_name)
>   {
> +	struct mux_lookup *m, *chosen = NULL;
> +	const char *dev_id = dev_name(dev);
> +	struct mux_chip *mux_chip;
> +
>   	/* look up via DT first */
>   	if (IS_ENABLED(CONFIG_OF) && dev->of_node)
>   		return of_mux_control_get(dev, mux_name);
>   
> -	return ERR_PTR(-ENODEV);
> +	/*
> +	 * For non DT we look up the provider in the static table typically
> +	 * provided by board setup code.
> +	 *
> +	 * If a match is found, the provider mux chip is looked up by name
> +	 * and a mux-control is requested using the table provided index.
> +	 */
> +	mutex_lock(&mux_lookup_lock);
> +	list_for_each_entry(m, &mux_lookup_list, list) {
> +		if (WARN_ON(!m->dev_id || !m->mux_name || !m->provider))
> +			continue;
> +
> +		if (strcmp(m->dev_id, dev_id) == 0 &&
> +		    strcmp(m->mux_name, mux_name) == 0) {
> +			chosen = m;
> +			break;
> +		}
> +	}
> +	mutex_unlock(&mux_lookup_lock);
> +
> +	if (!chosen)
> +		return ERR_PTR(-ENODEV);
> +
> +	mux_chip = mux_chip_get_by_name(chosen->provider);
> +	if (!mux_chip)
> +		return ERR_PTR(-EPROBE_DEFER);
> +
> +	if (chosen->index >= mux_chip->controllers) {
> +		dev_err(dev, "Mux lookup table index out of bounds %u >= %u\n",
> +			chosen->index, mux_chip->controllers);
> +		put_device(&mux_chip->dev);
> +		return ERR_PTR(-EINVAL);
> +	}
> +
> +	return &mux_chip->mux[chosen->index];
>   }
>   EXPORT_SYMBOL_GPL(mux_control_get);
>   
> diff --git a/include/linux/mux/consumer.h b/include/linux/mux/consumer.h
> index ea96d4c82be7..912dd48a3a5d 100644
> --- a/include/linux/mux/consumer.h
> +++ b/include/linux/mux/consumer.h
> @@ -18,6 +18,17 @@
>   struct device;
>   struct mux_control;
>   
> +struct mux_lookup {
> +	struct list_head list;
> +	const char *provider;
> +	unsigned int index;
> +	const char *dev_id;
> +	const char *mux_name;
> +};
> +
> +void mux_add_table(struct mux_lookup *table, size_t num);
> +void mux_remove_table(struct mux_lookup *table, size_t num);
> +
>   unsigned int mux_control_states(struct mux_control *mux);
>   int __must_check mux_control_select(struct mux_control *mux,
>   				    unsigned int state);

[toc] | [prev] | [next] | [standalone]

#1726107 — Re: [PATCH 02/11] mux: core: Add support for getting a mux controller on a non DT platform

FromHans de Goede <hdegoede@redhat.com>
Date2017-09-04 16:30 +0200
SubjectRe: [PATCH 02/11] mux: core: Add support for getting a mux controller on a non DT platform
Message-ID<um0xY-6Aw-3@gated-at.bofh.it>
In reply to#1725553
Hi,

On 02-09-17 21:13, sathya wrote:
> Hi,
> 
> 
> On 09/01/2017 02:48 PM, Hans de Goede wrote:
>> On non DT platforms we cannot get the mux_chip by pnode. Other subsystems
>> (regulator, clock, pwm) have the same problem and solve this by allowing
>> platform / board-setup code to add entries to a lookup table and then use
>> this table to look things up.
>>
>> This commit adds support for getting a mux controller on a non DT platform
>> following this pattern. It is based on a simplified version of the pwm
>> subsys lookup code, the dev_id and mux_name parts of a lookup table entry
>> are mandatory in the mux-core implementation.
>>
>> Signed-off-by: Hans de Goede <hdegoede@redhat.com>
>> ---
>>   drivers/mux/core.c           | 96 +++++++++++++++++++++++++++++++++++++++++++-
>>   include/linux/mux/consumer.h | 11 +++++
>>   2 files changed, 106 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/mux/core.c b/drivers/mux/core.c
>> index 6142493c327b..8864cc745506 100644
>> --- a/drivers/mux/core.c
>> +++ b/drivers/mux/core.c
>> @@ -24,6 +24,9 @@
>>   #include <linux/of_platform.h>
>>   #include <linux/slab.h>
>> +static DEFINE_MUTEX(mux_lookup_lock);
>> +static LIST_HEAD(mux_lookup_list);
>> +
>>   /*
>>    * The idle-as-is "state" is not an actual state that may be selected, it
>>    * only implies that the state should not be changed. So, use that state
>> @@ -408,6 +411,23 @@ int mux_control_deselect(struct mux_control *mux)
>>   }
>>   EXPORT_SYMBOL_GPL(mux_control_deselect);
>> +static int parent_name_match(struct device *dev, const void *data)
>> +{
>> +    const char *parent_name = dev_name(dev->parent);
> Device name usually contains id section (devname.id). Did you take this into consideration ?
>> +    const char *name = data;
>> +
>> +    return strcmp(parent_name, name) == 0;
>> +}
>> +
>> +static struct mux_chip *mux_chip_get_by_name(const char *name)
>> +{
>> +    struct device *dev;
>> +
>> +    dev = class_find_device(&mux_class, NULL, name, parent_name_match);
>> +
>> +    return dev ? to_mux_chip(dev) : NULL;
>> +}
>> +
>>   static int of_dev_node_match(struct device *dev, const void *data)
>>   {
>>       return dev->of_node == data;
>> @@ -479,6 +499,42 @@ static struct mux_control *of_mux_control_get(struct device *dev,
>>   }
>>   /**
>> + * mux_add_table() - register PWM device consumers
> PWM -> MUX

Thank you, good catch, fixed for v2.

Regards,

Hans


>> + * @table: array of consumers to register
>> + * @num: number of consumers in table
>> + */
>> +void mux_add_table(struct mux_lookup *table, size_t num)
>> +{
>> +    mutex_lock(&mux_lookup_lock);
>> +
>> +    while (num--) {
>> +        list_add_tail(&table->list, &mux_lookup_list);
>> +        table++;
>> +    }
>> +
>> +    mutex_unlock(&mux_lookup_lock);
>> +}
>> +EXPORT_SYMBOL_GPL(mux_add_table);
>> +
>> +/**
>> + * mux_remove_table() - unregister PWM device consumers
>> + * @table: array of consumers to unregister
>> + * @num: number of consumers in table
>> + */
>> +void mux_remove_table(struct mux_lookup *table, size_t num)
>> +{
>> +    mutex_lock(&mux_lookup_lock);
>> +
>> +    while (num--) {
>> +        list_del(&table->list);
>> +        table++;
>> +    }
>> +
>> +    mutex_unlock(&mux_lookup_lock);
>> +}
>> +EXPORT_SYMBOL_GPL(mux_remove_table);
>> +
>> +/**
>>    * mux_control_get() - Get the mux-control for a device.
>>    * @dev: The device that needs a mux-control.
>>    * @mux_name: The name identifying the mux-control.
>> @@ -487,11 +543,49 @@ static struct mux_control *of_mux_control_get(struct device *dev,
>>    */
>>   struct mux_control *mux_control_get(struct device *dev, const char *mux_name)
>>   {
>> +    struct mux_lookup *m, *chosen = NULL;
>> +    const char *dev_id = dev_name(dev);
>> +    struct mux_chip *mux_chip;
>> +
>>       /* look up via DT first */
>>       if (IS_ENABLED(CONFIG_OF) && dev->of_node)
>>           return of_mux_control_get(dev, mux_name);
>> -    return ERR_PTR(-ENODEV);
>> +    /*
>> +     * For non DT we look up the provider in the static table typically
>> +     * provided by board setup code.
>> +     *
>> +     * If a match is found, the provider mux chip is looked up by name
>> +     * and a mux-control is requested using the table provided index.
>> +     */
>> +    mutex_lock(&mux_lookup_lock);
>> +    list_for_each_entry(m, &mux_lookup_list, list) {
>> +        if (WARN_ON(!m->dev_id || !m->mux_name || !m->provider))
>> +            continue;
>> +
>> +        if (strcmp(m->dev_id, dev_id) == 0 &&
>> +            strcmp(m->mux_name, mux_name) == 0) {
>> +            chosen = m;
>> +            break;
>> +        }
>> +    }
>> +    mutex_unlock(&mux_lookup_lock);
>> +
>> +    if (!chosen)
>> +        return ERR_PTR(-ENODEV);
>> +
>> +    mux_chip = mux_chip_get_by_name(chosen->provider);
>> +    if (!mux_chip)
>> +        return ERR_PTR(-EPROBE_DEFER);
>> +
>> +    if (chosen->index >= mux_chip->controllers) {
>> +        dev_err(dev, "Mux lookup table index out of bounds %u >= %u\n",
>> +            chosen->index, mux_chip->controllers);
>> +        put_device(&mux_chip->dev);
>> +        return ERR_PTR(-EINVAL);
>> +    }
>> +
>> +    return &mux_chip->mux[chosen->index];
>>   }
>>   EXPORT_SYMBOL_GPL(mux_control_get);
>> diff --git a/include/linux/mux/consumer.h b/include/linux/mux/consumer.h
>> index ea96d4c82be7..912dd48a3a5d 100644
>> --- a/include/linux/mux/consumer.h
>> +++ b/include/linux/mux/consumer.h
>> @@ -18,6 +18,17 @@
>>   struct device;
>>   struct mux_control;
>> +struct mux_lookup {
>> +    struct list_head list;
>> +    const char *provider;
>> +    unsigned int index;
>> +    const char *dev_id;
>> +    const char *mux_name;
>> +};
>> +
>> +void mux_add_table(struct mux_lookup *table, size_t num);
>> +void mux_remove_table(struct mux_lookup *table, size_t num);
>> +
>>   unsigned int mux_control_states(struct mux_control *mux);
>>   int __must_check mux_control_select(struct mux_control *mux,
>>                       unsigned int state);
>

[toc] | [prev] | [next] | [standalone]

#1726613 — Re: [PATCH 02/11] mux: core: Add support for getting a mux controller on a non DT platform

FromHans de Goede <hdegoede@redhat.com>
Date2017-09-05 13:00 +0200
SubjectRe: [PATCH 02/11] mux: core: Add support for getting a mux controller on a non DT platform
Message-ID<umjKi-1Bt-19@gated-at.bofh.it>
In reply to#1725383
Hi,

On 04-09-17 13:19, Peter Rosin wrote:
> Hi!
> 
> Some comments inline...
> 
> On 2017-09-01 23:48, Hans de Goede wrote:
>> On non DT platforms we cannot get the mux_chip by pnode. Other subsystems
>> (regulator, clock, pwm) have the same problem and solve this by allowing
>> platform / board-setup code to add entries to a lookup table and then use
>> this table to look things up.
>>
>> This commit adds support for getting a mux controller on a non DT platform
>> following this pattern. It is based on a simplified version of the pwm
>> subsys lookup code, the dev_id and mux_name parts of a lookup table entry
>> are mandatory in the mux-core implementation.
>>
>> Signed-off-by: Hans de Goede <hdegoede@redhat.com>
>> ---
>>   drivers/mux/core.c           | 96 +++++++++++++++++++++++++++++++++++++++++++-
>>   include/linux/mux/consumer.h | 11 +++++
>>   2 files changed, 106 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/mux/core.c b/drivers/mux/core.c
>> index 6142493c327b..8864cc745506 100644
>> --- a/drivers/mux/core.c
>> +++ b/drivers/mux/core.c
>> @@ -24,6 +24,9 @@
>>   #include <linux/of_platform.h>
>>   #include <linux/slab.h>
>>   
>> +static DEFINE_MUTEX(mux_lookup_lock);
>> +static LIST_HEAD(mux_lookup_list);
>> +
>>   /*
>>    * The idle-as-is "state" is not an actual state that may be selected, it
>>    * only implies that the state should not be changed. So, use that state
>> @@ -408,6 +411,23 @@ int mux_control_deselect(struct mux_control *mux)
>>   }
>>   EXPORT_SYMBOL_GPL(mux_control_deselect);
>>   
>> +static int parent_name_match(struct device *dev, const void *data)
>> +{
>> +	const char *parent_name = dev_name(dev->parent);
>> +	const char *name = data;
>> +
>> +	return strcmp(parent_name, name) == 0;
>> +}
>> +
>> +static struct mux_chip *mux_chip_get_by_name(const char *name)
>> +{
>> +	struct device *dev;
>> +
>> +	dev = class_find_device(&mux_class, NULL, name, parent_name_match);
>> +
>> +	return dev ? to_mux_chip(dev) : NULL;
>> +}
>> +
>>   static int of_dev_node_match(struct device *dev, const void *data)
>>   {
>>   	return dev->of_node == data;
>> @@ -479,6 +499,42 @@ static struct mux_control *of_mux_control_get(struct device *dev,
>>   }
>>   
>>   /**
>> + * mux_add_table() - register PWM device consumers
> 
> register mux controllers (because you are not registering consumers, right?
> someone is registering controllers so that they can be found by consumers?)


Actually what is being registered is a "consumer to mux-controller mapping",
I will update the kernel-doc comments to use that everywhere.

> 
>> + * @table: array of consumers to register
>> + * @num: number of consumers in table
> 
> controllers?

mappings :)

>> + */
>> +void mux_add_table(struct mux_lookup *table, size_t num)
>> +{
>> +	mutex_lock(&mux_lookup_lock);
>> +
>> +	while (num--) {
>> +		list_add_tail(&table->list, &mux_lookup_list);
>> +		table++;
>> +	}
> 
> I prefer
> 
> 	for (; num--; table++)
> 		list_add_tail(&table->list, &mux_lookup_list);

Sure, works for me.

>> +
>> +	mutex_unlock(&mux_lookup_lock);
>> +}
>> +EXPORT_SYMBOL_GPL(mux_add_table);
>> +
>> +/**
>> + * mux_remove_table() - unregister PWM device consumers
> 
> unregister mux controllers(?)
> 
>> + * @table: array of consumers to unregister
>> + * @num: number of consumers in table
> 
> controllers?
> 
>> + */
>> +void mux_remove_table(struct mux_lookup *table, size_t num)
>> +{
>> +	mutex_lock(&mux_lookup_lock);
>> +
>> +	while (num--) {
>> +		list_del(&table->list);
>> +		table++;
>> +	}
> 
> for() loop here as well.

Ack.

>> +
>> +	mutex_unlock(&mux_lookup_lock);
>> +}
>> +EXPORT_SYMBOL_GPL(mux_remove_table);
>> +
>> +/**
>>    * mux_control_get() - Get the mux-control for a device.
>>    * @dev: The device that needs a mux-control.
>>    * @mux_name: The name identifying the mux-control.
>> @@ -487,11 +543,49 @@ static struct mux_control *of_mux_control_get(struct device *dev,
>>    */
>>   struct mux_control *mux_control_get(struct device *dev, const char *mux_name)
>>   {
>> +	struct mux_lookup *m, *chosen = NULL;
>> +	const char *dev_id = dev_name(dev);
>> +	struct mux_chip *mux_chip;
>> +
>>   	/* look up via DT first */
>>   	if (IS_ENABLED(CONFIG_OF) && dev->of_node)
>>   		return of_mux_control_get(dev, mux_name);
>>   
>> -	return ERR_PTR(-ENODEV);
>> +	/*
>> +	 * For non DT we look up the provider in the static table typically
>> +	 * provided by board setup code.
>> +	 *
>> +	 * If a match is found, the provider mux chip is looked up by name
>> +	 * and a mux-control is requested using the table provided index.
>> +	 */
>> +	mutex_lock(&mux_lookup_lock);
>> +	list_for_each_entry(m, &mux_lookup_list, list) {
>> +		if (WARN_ON(!m->dev_id || !m->mux_name || !m->provider))
>> +			continue;
>> +
>> +		if (strcmp(m->dev_id, dev_id) == 0 &&
>> +		    strcmp(m->mux_name, mux_name) == 0) {
> 
> I want the below format (with ! instead of == 0 and the brace on the next line
> when the condition has a line break):

Ok, I think checkpatch is going to not like that "{" there, but
I'm fine with putting it there.

> 		if (!strcmp(m->dev_id, dev_id) &&
> 		    !strcmp(m->mux_name, mux_name))
> 		{
> 
>> +			chosen = m;
>> +			break;
>> +		}
>> +	}
>> +	mutex_unlock(&mux_lookup_lock);
>> +
>> +	if (!chosen)
>> +		return ERR_PTR(-ENODEV);
>> +
>> +	mux_chip = mux_chip_get_by_name(chosen->provider);
>> +	if (!mux_chip)
>> +		return ERR_PTR(-EPROBE_DEFER);
>> +
>> +	if (chosen->index >= mux_chip->controllers) {
>> +		dev_err(dev, "Mux lookup table index out of bounds %u >= %u\n",
>> +			chosen->index, mux_chip->controllers);
>> +		put_device(&mux_chip->dev);
>> +		return ERR_PTR(-EINVAL);
>> +	}
>> +
>> +	return &mux_chip->mux[chosen->index];
>>   }
>>   EXPORT_SYMBOL_GPL(mux_control_get);
>>   
>> diff --git a/include/linux/mux/consumer.h b/include/linux/mux/consumer.h
>> index ea96d4c82be7..912dd48a3a5d 100644
>> --- a/include/linux/mux/consumer.h
>> +++ b/include/linux/mux/consumer.h
>> @@ -18,6 +18,17 @@
>>   struct device;
>>   struct mux_control;
>>   
> 
> I want a kernel-doc comment here, describing the structure.

Ok.

>> +struct mux_lookup {
>> +	struct list_head list;
>> +	const char *provider;
>> +	unsigned int index;
>> +	const char *dev_id;
>> +	const char *mux_name;
>> +};
>> +
>> +void mux_add_table(struct mux_lookup *table, size_t num);
>> +void mux_remove_table(struct mux_lookup *table, size_t num);
>> +
> 
> I'm not sure if consumer.h is the right place for this, but it can
> be moved when I think of something better. Which I can't for the
> moment...
> 
>>   unsigned int mux_control_states(struct mux_control *mux);
>>   int __must_check mux_control_select(struct mux_control *mux,
>>   				    unsigned int state);
>>
> 

I will address all comments for v2 of series.

Regards,

Hans

[toc] | [prev] | [next] | [standalone]

#1726610 — Re: [PATCH 00/11] mux/typec: Add USB / TypeC mux drivers and hook them up on some x86 systems

FromHans de Goede <hdegoede@redhat.com>
Date2017-09-05 13:00 +0200
SubjectRe: [PATCH 00/11] mux/typec: Add USB / TypeC mux drivers and hook them up on some x86 systems
Message-ID<umjKi-1Bt-9@gated-at.bofh.it>
In reply to#1725368
Hi Peter,

On 04-09-17 13:18, Peter Rosin wrote:
> On 2017-09-01 23:48, Hans de Goede wrote:
>> Hi All,
>>
>> This series consists of 4 parts:
>>
>> 1) Core mux changes to add support for getting mux-controllers on
>>     non DT platforms and to add some standardised state values for USB
>>
>> 2) Add Intel CHT USB mux and Pericom-PI3USB30532 Type-C mux drivers
>>
>> 3) Hookup the Intel CHT USB mux driver for CHT devices without Type-C
>>
>> 4) Hookup both the Intel CHT USB mux and Pericom-PI3USB30532 Type-C mux
>>     drivers to the fusb302 Type-C port controller found on some CHT x86 devs
>>
>> Please review, or in the case of the drivers/mux changes (which are a dep
>> for some of the other patches) merge if the patches are ready.
> 
> Hi Hans,
> 
> I see commonalities with this and the below patch seriess from Stephen
> Boyd [added to Cc].
> 
> https://lkml.org/lkml/2017/7/11/696 [PATCH 0/3] USB Mux support for Chipidea
> https://lkml.org/lkml/2017/7/14/654 [PATCH v2 0/3] USB Mux support for Chipidea

Interesting, it seems that Stephen and I are trying to use the mux-framework
for identical (device/host selection) purposes, except that in some cases
I also gave a Type-C cross-switch to deal with.

I noticed this discussion in the thread:

"""
 > +		usb-switch-states = <0>, <1>;

I don't see the need for usb-switch-states? Just assume states 0/1 and
if someone later needs some other states, make them add a property that
overrides the defaults. Just document that 0 is host and 1 is device.
"""

I think it makes sense to just agree on fixed "state" values for
certain use-cases, as done in my
"mux: consumer.h: Add MUX_USB_* state constant defines"
patch. This means that the "state" register in the hardware and the
state as passed to mux_control_select() may not map 1:1, but that can
easily be mapped in the driver and it allows inter-changeble (compatible)
mux drivers for some common mux use-cases such as an USB device/host
mux.

Edit: Ah I see you already suggest this below, good :)

> Stephen had a patch that added mux_control_get_optional() that I think
> could be of use for this series?

Ack, I think that would be useful. If you plan to merge Stephen's
patch for this, please give me a link to a git repo with that merged
then I will rebase my series on top.

> Anyway, there seems to be some interest in using the mux subsystem for
> handling the USB host/device role. I think the role-switch interface
> between the USB and mux subsystems should be done similarly, regardless
> of what particular USB driver needs to switch role using whatever
> particular mux driver. If at all possible. >
> The way I read it, the pi3usb30532 driver is the one adding the need to
> involve the DP states and the inverted bit.

Correct, the pi3usb30532 mux is a mux designed for Type-C ports, it
switches to 4 high-speed differential data-pairs the Type-C connector
has to one on: USB-controller(s), DisplayPort outputs on the GPU, or
a combination of both (2 data-pairs to each). It also takes into account
if the Type-C connector has been inserted normally or "upside-down",
which is where the inverted bits comes into play.

> Those things are not used by
> anything else, and I think it clutters up things for everybody when the
> weird needs of one driver "invades" the mux states needed to control the
> USB role. So, I would like USB role switching to have 2 states, device
> and host (0 and 1 is used by the above chipidea code). And then the USB
> switch can of course be idle, which is best represented with one of
> MUX_IDLE_DISCONNECT, MUX_IDLE_AS_IS or one of the modes depending on if
> the USB switch can or can't disconnect (or other considerations). Now,
> you can't explicitly set the idle state using mux_control_select(), it
> can only be set implicitly using mux_control_deselect(), so that will
> require some changes in the consumer logic.
> 
> Regarding the pi3usb30532 driver, I think the above is in fact also
> better since the "swap bit" means something totally different when the
> switch is "open" (if I read the datasheet correctly >
> I.e. PI3USB30532_CONF_OPEN | PI3USB30532_CONF_SWAP is not sane,

The only "datasheet" I could find is "PI3USB30532-PI3USB31532_App_Type-C application Note_Rev.A.pdf"
and that says the swap bits does not do anything when the switch is "open"
but that does not matter for the rest of the discussion, I do agree
that mapping deselected to open makes sense.

> and that
> would just go away completely if the driver implemented MUX_IDLE_DISCONNECT
> as PI3USB30532_CONF_OPEN and removed the possibility to explicitly set the
> "open" state with mux_control_select().
> 
> So, I think the generic USB role switch interface should be something like:
> 
> #define MUX_USB_DEVICE		(0) /* USB device mode */
> #define MUX_USB_HOST		(1) /* USB host mode */

Those 2 work for me.

> And then the pi3usb30532 driver can extend that:
> 
> #define MUX_PI3USB30352_HOST_AND_DP_SRC	(2)    /* USB host + 2 lanes Display Port */
> #define MUX_PI3USB30352_DP_SRC		(3)    /* 4 lanes Display Port source */
> #define MUX_PI3USB30352_POLARITY_INV	BIT(2) /* Polarity inverted bit */
> #define MUX_PI3USB30352_STATES		(8)

Erm, I would like to keep mux-driver specific knowledge out of the
tcpm code and there might be more Type-C mux drivers in the future,
how about:

#define MUX_TYPEC_POLARITY_INV    BIT(0)   /* Polarity inverted bit */
#define MUX_TYPEC_DEVICE          (0 << 1) /* USB device mode */
#define MUX_TYPEC_HOST            (1 << 1) /* USB host mode */
#define MUX_TYPEC_HOST_AND_DP_SRC (2 << 1) /* USB host + 2 lanes Display Port */
#define MUX_TYPEC_DP_SRC          (3 << 1) /* 4 lanes Display Port source */
#define MUX_TYPEC_STATES          (4 << 1)

?

> Another idea is to expose the inverted bit as a separate mux controller,
> but I suspect that you're not too keen on operating three muxes in the
> tcpm driver...

That won't be pretty I think, so thanks but no thanks :)

> BTW, I don't think these USB defines belong in mux/consumer.h, please add
> them in a new include/linux/mux/usb.h file or something. And I'd like some
> MAINTAINER entry to cover the new mux drivers...

Ok, I will fix both for the next version

> I'll respond to the individual patches with nits etc, but it generally looks
> tidy, thank you for that!

Thank you for all the feedback and reviews.

Regards,

Hans

[toc] | [prev] | [standalone]

Page 2 of 2 — ← Prev page 1 [2]

Back to top | Article view | linux.kernel

csiph-web