On Mon, Sep 19, 2016 at 02:16:56PM +0300, Heikki Krogerus wrote:
> The purpose of USB Type-C connector class is to provide
> unified interface for the user space to get the status and
> basic information about USB Type-C connectors on a system,
> control over data role swapping, and when the port supports
> USB Power Delivery, also control over power role swapping
> and Alternate Modes.
>
> Reviewed-by: Guenter Roeck <[email protected]>
> Tested-by: Guenter Roeck <[email protected]>
> Signed-off-by: Heikki Krogerus <[email protected]>
> ---
> Documentation/ABI/testing/sysfs-class-typec | 218 ++++++
> Documentation/usb/typec.txt | 103 +++
> MAINTAINERS | 9 +
> drivers/usb/Kconfig | 2 +
> drivers/usb/Makefile | 2 +
> drivers/usb/typec/Kconfig | 7 +
> drivers/usb/typec/Makefile | 1 +
> drivers/usb/typec/typec.c | 1075 +++++++++++++++++++++++++++
> include/linux/usb/typec.h | 252 +++++++
> 9 files changed, 1669 insertions(+)
> create mode 100644 Documentation/ABI/testing/sysfs-class-typec
> create mode 100644 Documentation/usb/typec.txt
> create mode 100644 drivers/usb/typec/Kconfig
> create mode 100644 drivers/usb/typec/Makefile
> create mode 100644 drivers/usb/typec/typec.c
> create mode 100644 include/linux/usb/typec.h
Overall, this looks good, just a few minor comments that will require
you to respin this...
>
> diff --git a/Documentation/ABI/testing/sysfs-class-typec b/Documentation/ABI/testing/sysfs-class-typec
> new file mode 100644
> index 0000000..dcca6bd
> --- /dev/null
> +++ b/Documentation/ABI/testing/sysfs-class-typec
> @@ -0,0 +1,218 @@
> +USB Type-C port devices (eg. /sys/class/typec/usbc0/)
> +
> +What: /sys/class/typec/<port>/current_data_role
> +Date: June 2016
It's no longer June. I don't know why we have these dates, sorry, just
make it be December and you should be fine.
> --- /dev/null
> +++ b/Documentation/usb/typec.txt
We want to use .rst formats now, but this should be fine as-is for now.
> +static int sysfs_strmatch(const char * const *array, size_t n, const char *str)
> +{
> + const char *item;
> + int index;
> +
> + for (index = 0; index < n; index++) {
> + item = array[index];
> + if (!item)
> + break;
> + if (sysfs_streq(item, str))
> + return index;
> + }
> +
> + return -EINVAL;
> +}
should we make this a core sysfs function?
> +
> +/* ------------------------------------------------------------------------- */
> +/* Type-C Partners */
> +
> +static void typec_dev_release(struct device *dev)
> +{
> +}
Yeah, thanks to the in-kernel documentation, I now get to make fun of
you!!!!
Please, NEVER DO THIS EVER!!! You are trying to tell the kernel "hey,
shut up for your stupid warning about an empty release function!" Did
you ever think about _why_ I made the kernel warn about that? Don't
think you are smarter than the kernel, you will always loose that bet...
> +
> +static ssize_t partner_usb_pd_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct typec_partner *p = container_of(dev, struct typec_partner, dev);
> +
> + return sprintf(buf, "%d\n", p->usb_pd);
> +}
> +
> +static struct device_attribute dev_attr_partner_usb_pd = {
> + .attr = {
> + .name = "supports_usb_power_delivery",
> + .mode = S_IRUGO,
> + },
> + .show = partner_usb_pd_show,
> +};
DEVICE_ATTR_RO()?
> +
> +static ssize_t
> +partner_accessory_mode_show(struct device *dev, struct device_attribute *attr,
> + char *buf)
> +{
> + struct typec_partner *p = container_of(dev, struct typec_partner, dev);
> +
> + return sprintf(buf, "%s\n", typec_accessory_modes[p->accessory]);
> +}
> +
> +static struct device_attribute dev_attr_partner_accessory = {
> + .attr = {
> + .name = "accessory_mode",
> + .mode = S_IRUGO,
> + },
> + .show = partner_accessory_mode_show,
> +};
DEVICE_ATTR_RO()?
> +
> +static struct attribute *typec_partner_attrs[] = {
> + &dev_attr_partner_accessory.attr,
> + &dev_attr_partner_usb_pd.attr,
> + NULL
> +};
> +
> +static struct attribute_group typec_partner_group = {
> + .attrs = typec_partner_attrs,
> +};
> +
> +static const struct attribute_group *typec_partner_groups[] = {
> + &typec_partner_group,
> + NULL
> +};
ATTRIBUTE_GROUPS()?
> +
> +static struct device_type typec_partner_dev_type = {
> + .name = "typec_partner_device",
> + .groups = typec_partner_groups,
> + .release = typec_dev_release,
> +};
> +
> +static int
> +typec_add_partner(struct typec_port *port, struct typec_partner *partner)
> +{
> + struct device *dev = &partner->dev;
> + int ret;
> +
> + dev->class = &typec_class;
> + dev->parent = &port->dev;
> + dev->type = &typec_partner_dev_type;
> + dev_set_name(dev, "%s-partner", dev_name(&port->dev));
> +
> + ret = device_register(dev);
> + if (ret) {
> + put_device(dev);
> + return ret;
> + }
> +
> + port->partner = partner;
> + return 0;
> +}
> +
> +static void typec_remove_partner(struct typec_port *port)
> +{
> + WARN_ON(port->partner->alt_modes);
how can this happen? What would a user do if it does?
> + device_unregister(&port->partner->dev);
> +}
> +
> +/* ------------------------------------------------------------------------- */
> +/* Type-C Cable Plugs */
> +
> +static struct device_type typec_plug_dev_type = {
> + .name = "typec_plug_device",
> + .release = typec_dev_release,
> +};
> +
> +static int
> +typec_add_plug(struct typec_port *port, struct typec_plug *plug)
> +{
> + struct device *dev = &plug->dev;
> + char name[8];
> + int ret;
> +
> + sprintf(name, "plug%d", plug->index);
> +
> + dev->class = &typec_class;
> + dev->parent = &port->cable->dev;
> + dev->type = &typec_plug_dev_type;
> + dev_set_name(dev, "%s-%s", dev_name(&port->dev), name);
> +
> + ret = device_register(dev);
> + if (ret) {
> + put_device(dev);
> + return ret;
> + }
> +
> + return 0;
> +}
> +
> +static void typec_remove_plug(struct typec_plug *plug)
> +{
> + WARN_ON(plug->alt_modes);
again, what can someone do with this?
> + device_unregister(&plug->dev);
> +}
> +
> +/* Type-C Cables */
> +
> +static ssize_t
> +cable_active_show(struct device *dev, struct device_attribute *attr, char *buf)
> +{
> + struct typec_cable *cable = container_of(dev, struct typec_cable, dev);
> +
> + return sprintf(buf, "%d\n", cable->active);
> +}
> +
> +static struct device_attribute dev_attr_cable_active = {
> + .attr = {
> + .name = "active",
> + .mode = S_IRUGO,
> + },
> + .show = cable_active_show,
> +};
DEVICE_ATTR_RO()
> +
> +static ssize_t cable_usb_pd_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct typec_cable *cable = container_of(dev, struct typec_cable, dev);
> +
> + return sprintf(buf, "%d\n", cable->usb_pd);
> +}
> +
> +static struct device_attribute dev_attr_cable_usb_pd = {
> + .attr = {
> + .name = "supports_usb_power_delivery",
> + .mode = S_IRUGO,
> + },
> + .show = cable_usb_pd_show,
> +};
DEVICE_ATTR_RO()
> +
> +static const char * const typec_plug_types[] = {
> + [USB_PLUG_NONE] = "unknown",
> + [USB_PLUG_TYPE_A] = "Type-A",
> + [USB_PLUG_TYPE_B] = "Type-B",
> + [USB_PLUG_TYPE_C] = "Type-C",
> + [USB_PLUG_CAPTIVE] = "Captive",
> +};
> +
> +static ssize_t cable_plug_type_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct typec_cable *cable = container_of(dev, struct typec_cable, dev);
> +
> + return sprintf(buf, "%s\n", typec_plug_types[cable->type]);
> +}
> +
> +static struct device_attribute dev_attr_plug_type = {
> + .attr = {
> + .name = "plug_type",
> + .mode = S_IRUGO,
> + },
> + .show = cable_plug_type_show,
> +};
And so on...
> +
> +static struct attribute *typec_cable_attrs[] = {
> + &dev_attr_cable_active.attr,
> + &dev_attr_cable_usb_pd.attr,
> + &dev_attr_plug_type.attr,
> + NULL
> +};
> +
> +static struct attribute_group typec_cable_group = {
> + .attrs = typec_cable_attrs,
> +};
> +
> +static const struct attribute_group *typec_cable_groups[] = {
> + &typec_cable_group,
> + NULL
> +};
ATTRIBUTE_GROUPS()?
> +
> +static struct device_type typec_cable_dev_type = {
> + .name = "typec_cable_device",
> + .groups = typec_cable_groups,
> + .release = typec_dev_release,
> +};
> +
> +static int typec_add_cable(struct typec_port *port, struct typec_cable *cable)
> +{
> + struct device *dev = &cable->dev;
> + int ret;
> +
> + dev->class = &typec_class;
> + dev->parent = &port->dev;
> + dev->type = &typec_cable_dev_type;
> + dev_set_name(dev, "%s-cable", dev_name(&port->dev));
> +
> + ret = device_register(dev);
> + if (ret) {
> + put_device(dev);
> + return ret;
> + }
> +
> + /* Plug1 */
> + if (!cable->usb_pd)
> + return 0;
> +
> + cable->plug[0].index = 1;
> + ret = typec_add_plug(port, &cable->plug[0]);
> + if (ret) {
> + device_unregister(dev);
> + return ret;
> + }
> +
> + /* Plug2 */
> + if (!cable->active || !cable->sop_pp_controller)
> + return 0;
> +
> + cable->plug[1].index = 2;
> + ret = typec_add_plug(port, &cable->plug[1]);
> + if (ret) {
> + typec_remove_plug(&cable->plug[0]);
> + device_unregister(dev);
> + return ret;
> + }
> +
> + port->cable = cable;
> + return 0;
> +}
> +
> +static void typec_remove_cable(struct typec_port *port)
> +{
> + if (port->cable->active) {
> + typec_remove_plug(&port->cable->plug[0]);
> + if (port->cable->sop_pp_controller)
> + typec_remove_plug(&port->cable->plug[1]);
> + }
> + device_unregister(&port->cable->dev);
> +}
> +
> +/* ------------------------------------------------------------------------- */
> +/* API for the port drivers */
> +
> +static void typec_init_roles(struct typec_port *port)
> +{
> + if (port->prefer_role < 0)
> + return;
> +
> + if (port->prefer_role == TYPEC_SOURCE) {
> + port->data_role = TYPEC_HOST;
> + port->pwr_role = TYPEC_SOURCE;
> + port->vconn_role = TYPEC_SOURCE;
> + } else {
> + /* Device mode as default also by default with DRP ports */
> + port->data_role = TYPEC_DEVICE;
> + port->pwr_role = TYPEC_SINK;
> + port->vconn_role = TYPEC_SINK;
> + }
> +}
> +
> +int typec_connect(struct typec_port *port, struct typec_connection *con)
> +{
> + int ret;
> +
> + if (!con->partner && !con->cable)
> + return -EINVAL;
> +
> + port->connected = 1;
> + port->data_role = con->data_role;
> + port->pwr_role = con->pwr_role;
> + port->vconn_role = con->vconn_role;
> + port->pwr_opmode = con->pwr_opmode;
> +
> + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
This worries me. Who is listening for it? What will you do with it?
Shouldn't you just poll on an attribute file instead?
> +
> + if (con->cable) {
> + ret = typec_add_cable(port, con->cable);
> + if (ret)
> + return ret;
> + }
> +
> + if (con->partner) {
> + ret = typec_add_partner(port, con->partner);
> + if (ret) {
> + if (con->cable)
> + typec_remove_cable(port);
> + return ret;
> + }
> + }
> +
> + return 0;
> +}
> +EXPORT_SYMBOL_GPL(typec_connect);
> +
> +void typec_disconnect(struct typec_port *port)
> +{
> + if (port->partner)
> + typec_remove_partner(port);
> +
> + if (port->cable)
> + typec_remove_cable(port);
> +
> + port->connected = 0;
> + port->partner = NULL;
> + port->cable = NULL;
> +
> + port->pwr_opmode = TYPEC_PWR_MODE_USB;
> +
> + typec_init_roles(port);
> +
> + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
> +}
> +EXPORT_SYMBOL_GPL(typec_disconnect);
> +
> +/* --------------------------------------- */
> +/* Driver callbacks to report role updates */
> +
> +void typec_set_data_role(struct typec_port *port, enum typec_data_role role)
> +{
> + port->data_role = role;
> + sysfs_notify(&port->dev.kobj, NULL, "current_data_role");
> +}
> +EXPORT_SYMBOL(typec_set_data_role);
EXPORT_SYMBOL_GPL() everywhere please, don't mix and match for no good
reason.
> +static void typec_init_modes(struct typec_altmode *alt, int is_port)
> +{
> + struct typec_mode *mode = alt->modes;
> + int i;
> +
> + for (i = 0; i < alt->n_modes; i++, mode++) {
> + mode->alt_mode = alt;
> + mode->index = i;
> + sprintf(mode->group_name, "mode%d", i);
> +
> + sysfs_attr_init(&mode->vdo_attr.attr);
> + mode->vdo_attr.attr.name = "vdo";
> + mode->vdo_attr.attr.mode = S_IRUGO;
> + mode->vdo_attr.show = typec_altmode_vdo_show;
> +
> + sysfs_attr_init(&mode->desc_attr.attr);
> + mode->desc_attr.attr.name = "description";
> + mode->desc_attr.attr.mode = S_IRUGO;
> + mode->desc_attr.show = typec_altmode_desc_show;
> +
> + sysfs_attr_init(&mode->active_attr.attr);
> + mode->active_attr.attr.name = "active";
> + mode->active_attr.attr.mode = S_IWUSR | S_IRUGO;
> + mode->active_attr.show = typec_altmode_active_show;
> + mode->active_attr.store = typec_altmode_active_store;
> +
> + mode->attrs[0] = &mode->vdo_attr.attr;
> + mode->attrs[1] = &mode->desc_attr.attr;
> + mode->attrs[2] = &mode->active_attr.attr;
> +
> + /* With ports, list the roles that the mode is supported with */
> + if (is_port) {
> + sysfs_attr_init(&mode->roles_attr.attr);
> + mode->roles_attr.attr.name = "supported_roles";
> + mode->roles_attr.attr.mode = S_IRUGO;
> + mode->roles_attr.show = typec_altmode_roles_show;
> +
> + mode->attrs[3] = &mode->roles_attr.attr;
> + }
> +
> + mode->group.attrs = mode->attrs;
> + mode->group.name = mode->group_name;
> +
> + alt->mode_groups[i] = &mode->group;
> + }
> +}
Ugh, dynamic attributes on the fly? why? Why not just use the callback
to determine if the attribute should be shown or not? Much simpler and
you don't have to clean up after yourself. Are you sure you cleaned up
properly from these?
> +static ssize_t current_power_role_show(struct device *dev,
> + struct device_attribute *attr, char *buf)
> +{
> + struct typec_port *port = to_typec_port(dev);
> +
> + return sprintf(buf, "%s\n", typec_roles[port->pwr_role]);
> +}
> +static DEVICE_ATTR_RW(current_power_role);
Nice, see, you use the macros here, be consistant please...
> +static const struct attribute_group typec_group = {
> + .attrs = typec_attrs,
> +};
> +
> +static const struct attribute_group *typec_groups[] = {
> + &typec_group,
> + NULL,
> +};
ATTRIBUTE_GROUP() please.
> +static int __init typec_init(void)
> +{
> + return class_register(&typec_class);
> +}
> +subsys_initcall(typec_init);
> +
> +static void __exit typec_exit(void)
> +{
> + class_unregister(&typec_class);
You forgot to clean up your idr :(
thanks,
greg k-h
Hi Greg,
On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
> On Mon, Sep 19, 2016 at 02:16:56PM +0300, Heikki Krogerus wrote:
> > The purpose of USB Type-C connector class is to provide
> > unified interface for the user space to get the status and
> > basic information about USB Type-C connectors on a system,
> > control over data role swapping, and when the port supports
> > USB Power Delivery, also control over power role swapping
> > and Alternate Modes.
> >
> > Reviewed-by: Guenter Roeck <[email protected]>
> > Tested-by: Guenter Roeck <[email protected]>
> > Signed-off-by: Heikki Krogerus <[email protected]>
> > ---
> > Documentation/ABI/testing/sysfs-class-typec | 218 ++++++
> > Documentation/usb/typec.txt | 103 +++
> > MAINTAINERS | 9 +
> > drivers/usb/Kconfig | 2 +
> > drivers/usb/Makefile | 2 +
> > drivers/usb/typec/Kconfig | 7 +
> > drivers/usb/typec/Makefile | 1 +
> > drivers/usb/typec/typec.c | 1075 +++++++++++++++++++++++++++
> > include/linux/usb/typec.h | 252 +++++++
> > 9 files changed, 1669 insertions(+)
> > create mode 100644 Documentation/ABI/testing/sysfs-class-typec
> > create mode 100644 Documentation/usb/typec.txt
> > create mode 100644 drivers/usb/typec/Kconfig
> > create mode 100644 drivers/usb/typec/Makefile
> > create mode 100644 drivers/usb/typec/typec.c
> > create mode 100644 include/linux/usb/typec.h
>
> Overall, this looks good, just a few minor comments that will require
> you to respin this...
>
>
> >
> > diff --git a/Documentation/ABI/testing/sysfs-class-typec b/Documentation/ABI/testing/sysfs-class-typec
> > new file mode 100644
> > index 0000000..dcca6bd
> > --- /dev/null
> > +++ b/Documentation/ABI/testing/sysfs-class-typec
> > @@ -0,0 +1,218 @@
> > +USB Type-C port devices (eg. /sys/class/typec/usbc0/)
> > +
> > +What: /sys/class/typec/<port>/current_data_role
> > +Date: June 2016
>
> It's no longer June. I don't know why we have these dates, sorry, just
> make it be December and you should be fine.
OK
> > --- /dev/null
> > +++ b/Documentation/usb/typec.txt
>
> We want to use .rst formats now, but this should be fine as-is for now.
>
> > +static int sysfs_strmatch(const char * const *array, size_t n, const char *str)
> > +{
> > + const char *item;
> > + int index;
> > +
> > + for (index = 0; index < n; index++) {
> > + item = array[index];
> > + if (!item)
> > + break;
> > + if (sysfs_streq(item, str))
> > + return index;
> > + }
> > +
> > + return -EINVAL;
> > +}
>
> should we make this a core sysfs function?
I already have the patch. I was planning on proposing that separately
after this series. This turned out to be something that can be used in
many drivers, so I was thinking about proposing it for at least few
other drivers.
But if you prefer, I can also introduce the function alone as new
sysfs core function in this series, and just use it in this driver.
> > +
> > +/* ------------------------------------------------------------------------- */
> > +/* Type-C Partners */
> > +
> > +static void typec_dev_release(struct device *dev)
> > +{
> > +}
>
> Yeah, thanks to the in-kernel documentation, I now get to make fun of
> you!!!!
>
> Please, NEVER DO THIS EVER!!! You are trying to tell the kernel "hey,
> shut up for your stupid warning about an empty release function!" Did
> you ever think about _why_ I made the kernel warn about that? Don't
> think you are smarter than the kernel, you will always loose that bet...
Point taken.
> > +
> > +static ssize_t partner_usb_pd_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct typec_partner *p = container_of(dev, struct typec_partner, dev);
> > +
> > + return sprintf(buf, "%d\n", p->usb_pd);
> > +}
> > +
> > +static struct device_attribute dev_attr_partner_usb_pd = {
> > + .attr = {
> > + .name = "supports_usb_power_delivery",
> > + .mode = S_IRUGO,
> > + },
> > + .show = partner_usb_pd_show,
> > +};
>
> DEVICE_ATTR_RO()?
I'm using the same attribute names with different types of devises. It
felt more wrong to use shared functions for some of the attributes
(but not all), and try to identify the device type in them, then to
simply ignore the macros and name the functions how ever I wanted. At
least it made the driver look less messy for sure.
But I'll try change these.
> > +
> > +static ssize_t
> > +partner_accessory_mode_show(struct device *dev, struct device_attribute *attr,
> > + char *buf)
> > +{
> > + struct typec_partner *p = container_of(dev, struct typec_partner, dev);
> > +
> > + return sprintf(buf, "%s\n", typec_accessory_modes[p->accessory]);
> > +}
> > +
> > +static struct device_attribute dev_attr_partner_accessory = {
> > + .attr = {
> > + .name = "accessory_mode",
> > + .mode = S_IRUGO,
> > + },
> > + .show = partner_accessory_mode_show,
> > +};
>
> DEVICE_ATTR_RO()?
>
> > +
> > +static struct attribute *typec_partner_attrs[] = {
> > + &dev_attr_partner_accessory.attr,
> > + &dev_attr_partner_usb_pd.attr,
> > + NULL
> > +};
> > +
> > +static struct attribute_group typec_partner_group = {
> > + .attrs = typec_partner_attrs,
> > +};
> > +
> > +static const struct attribute_group *typec_partner_groups[] = {
> > + &typec_partner_group,
> > + NULL
> > +};
>
> ATTRIBUTE_GROUPS()?
>
>
> > +
> > +static struct device_type typec_partner_dev_type = {
> > + .name = "typec_partner_device",
> > + .groups = typec_partner_groups,
> > + .release = typec_dev_release,
> > +};
> > +
> > +static int
> > +typec_add_partner(struct typec_port *port, struct typec_partner *partner)
> > +{
> > + struct device *dev = &partner->dev;
> > + int ret;
> > +
> > + dev->class = &typec_class;
> > + dev->parent = &port->dev;
> > + dev->type = &typec_partner_dev_type;
> > + dev_set_name(dev, "%s-partner", dev_name(&port->dev));
> > +
> > + ret = device_register(dev);
> > + if (ret) {
> > + put_device(dev);
> > + return ret;
> > + }
> > +
> > + port->partner = partner;
> > + return 0;
> > +}
> > +
> > +static void typec_remove_partner(struct typec_port *port)
> > +{
> > + WARN_ON(port->partner->alt_modes);
>
> how can this happen? What would a user do if it does?
I'll get rid of that.
> > + device_unregister(&port->partner->dev);
> > +}
> > +
> > +/* ------------------------------------------------------------------------- */
> > +/* Type-C Cable Plugs */
> > +
> > +static struct device_type typec_plug_dev_type = {
> > + .name = "typec_plug_device",
> > + .release = typec_dev_release,
> > +};
> > +
> > +static int
> > +typec_add_plug(struct typec_port *port, struct typec_plug *plug)
> > +{
> > + struct device *dev = &plug->dev;
> > + char name[8];
> > + int ret;
> > +
> > + sprintf(name, "plug%d", plug->index);
> > +
> > + dev->class = &typec_class;
> > + dev->parent = &port->cable->dev;
> > + dev->type = &typec_plug_dev_type;
> > + dev_set_name(dev, "%s-%s", dev_name(&port->dev), name);
> > +
> > + ret = device_register(dev);
> > + if (ret) {
> > + put_device(dev);
> > + return ret;
> > + }
> > +
> > + return 0;
> > +}
> > +
> > +static void typec_remove_plug(struct typec_plug *plug)
> > +{
> > + WARN_ON(plug->alt_modes);
>
> again, what can someone do with this?
I'll get rid of that too.
> > + device_unregister(&plug->dev);
> > +}
> > +
> > +/* Type-C Cables */
> > +
> > +static ssize_t
> > +cable_active_show(struct device *dev, struct device_attribute *attr, char *buf)
> > +{
> > + struct typec_cable *cable = container_of(dev, struct typec_cable, dev);
> > +
> > + return sprintf(buf, "%d\n", cable->active);
> > +}
> > +
> > +static struct device_attribute dev_attr_cable_active = {
> > + .attr = {
> > + .name = "active",
> > + .mode = S_IRUGO,
> > + },
> > + .show = cable_active_show,
> > +};
>
> DEVICE_ATTR_RO()
>
> > +
> > +static ssize_t cable_usb_pd_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct typec_cable *cable = container_of(dev, struct typec_cable, dev);
> > +
> > + return sprintf(buf, "%d\n", cable->usb_pd);
> > +}
> > +
> > +static struct device_attribute dev_attr_cable_usb_pd = {
> > + .attr = {
> > + .name = "supports_usb_power_delivery",
> > + .mode = S_IRUGO,
> > + },
> > + .show = cable_usb_pd_show,
> > +};
>
> DEVICE_ATTR_RO()
>
> > +
> > +static const char * const typec_plug_types[] = {
> > + [USB_PLUG_NONE] = "unknown",
> > + [USB_PLUG_TYPE_A] = "Type-A",
> > + [USB_PLUG_TYPE_B] = "Type-B",
> > + [USB_PLUG_TYPE_C] = "Type-C",
> > + [USB_PLUG_CAPTIVE] = "Captive",
> > +};
> > +
> > +static ssize_t cable_plug_type_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct typec_cable *cable = container_of(dev, struct typec_cable, dev);
> > +
> > + return sprintf(buf, "%s\n", typec_plug_types[cable->type]);
> > +}
> > +
> > +static struct device_attribute dev_attr_plug_type = {
> > + .attr = {
> > + .name = "plug_type",
> > + .mode = S_IRUGO,
> > + },
> > + .show = cable_plug_type_show,
> > +};
>
> And so on...
>
>
> > +
> > +static struct attribute *typec_cable_attrs[] = {
> > + &dev_attr_cable_active.attr,
> > + &dev_attr_cable_usb_pd.attr,
> > + &dev_attr_plug_type.attr,
> > + NULL
> > +};
> > +
> > +static struct attribute_group typec_cable_group = {
> > + .attrs = typec_cable_attrs,
> > +};
> > +
> > +static const struct attribute_group *typec_cable_groups[] = {
> > + &typec_cable_group,
> > + NULL
> > +};
>
> ATTRIBUTE_GROUPS()?
>
>
> > +
> > +static struct device_type typec_cable_dev_type = {
> > + .name = "typec_cable_device",
> > + .groups = typec_cable_groups,
> > + .release = typec_dev_release,
> > +};
> > +
> > +static int typec_add_cable(struct typec_port *port, struct typec_cable *cable)
> > +{
> > + struct device *dev = &cable->dev;
> > + int ret;
> > +
> > + dev->class = &typec_class;
> > + dev->parent = &port->dev;
> > + dev->type = &typec_cable_dev_type;
> > + dev_set_name(dev, "%s-cable", dev_name(&port->dev));
> > +
> > + ret = device_register(dev);
> > + if (ret) {
> > + put_device(dev);
> > + return ret;
> > + }
> > +
> > + /* Plug1 */
> > + if (!cable->usb_pd)
> > + return 0;
> > +
> > + cable->plug[0].index = 1;
> > + ret = typec_add_plug(port, &cable->plug[0]);
> > + if (ret) {
> > + device_unregister(dev);
> > + return ret;
> > + }
> > +
> > + /* Plug2 */
> > + if (!cable->active || !cable->sop_pp_controller)
> > + return 0;
> > +
> > + cable->plug[1].index = 2;
> > + ret = typec_add_plug(port, &cable->plug[1]);
> > + if (ret) {
> > + typec_remove_plug(&cable->plug[0]);
> > + device_unregister(dev);
> > + return ret;
> > + }
> > +
> > + port->cable = cable;
> > + return 0;
> > +}
> > +
> > +static void typec_remove_cable(struct typec_port *port)
> > +{
> > + if (port->cable->active) {
> > + typec_remove_plug(&port->cable->plug[0]);
> > + if (port->cable->sop_pp_controller)
> > + typec_remove_plug(&port->cable->plug[1]);
> > + }
> > + device_unregister(&port->cable->dev);
> > +}
> > +
> > +/* ------------------------------------------------------------------------- */
> > +/* API for the port drivers */
> > +
> > +static void typec_init_roles(struct typec_port *port)
> > +{
> > + if (port->prefer_role < 0)
> > + return;
> > +
> > + if (port->prefer_role == TYPEC_SOURCE) {
> > + port->data_role = TYPEC_HOST;
> > + port->pwr_role = TYPEC_SOURCE;
> > + port->vconn_role = TYPEC_SOURCE;
> > + } else {
> > + /* Device mode as default also by default with DRP ports */
> > + port->data_role = TYPEC_DEVICE;
> > + port->pwr_role = TYPEC_SINK;
> > + port->vconn_role = TYPEC_SINK;
> > + }
> > +}
> > +
> > +int typec_connect(struct typec_port *port, struct typec_connection *con)
> > +{
> > + int ret;
> > +
> > + if (!con->partner && !con->cable)
> > + return -EINVAL;
> > +
> > + port->connected = 1;
> > + port->data_role = con->data_role;
> > + port->pwr_role = con->pwr_role;
> > + port->vconn_role = con->vconn_role;
> > + port->pwr_opmode = con->pwr_opmode;
> > +
> > + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
>
> This worries me. Who is listening for it? What will you do with it?
> Shouldn't you just poll on an attribute file instead?
Oliver! Did you need this or can we remove it?
I remember I removed the "connected" attribute because you did not see
any use for it at one point. I don't remember the reason exactly why?
> > +
> > + if (con->cable) {
> > + ret = typec_add_cable(port, con->cable);
> > + if (ret)
> > + return ret;
> > + }
> > +
> > + if (con->partner) {
> > + ret = typec_add_partner(port, con->partner);
> > + if (ret) {
> > + if (con->cable)
> > + typec_remove_cable(port);
> > + return ret;
> > + }
> > + }
> > +
> > + return 0;
> > +}
> > +EXPORT_SYMBOL_GPL(typec_connect);
> > +
> > +void typec_disconnect(struct typec_port *port)
> > +{
> > + if (port->partner)
> > + typec_remove_partner(port);
> > +
> > + if (port->cable)
> > + typec_remove_cable(port);
> > +
> > + port->connected = 0;
> > + port->partner = NULL;
> > + port->cable = NULL;
> > +
> > + port->pwr_opmode = TYPEC_PWR_MODE_USB;
> > +
> > + typec_init_roles(port);
> > +
> > + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
> > +}
> > +EXPORT_SYMBOL_GPL(typec_disconnect);
> > +
> > +/* --------------------------------------- */
> > +/* Driver callbacks to report role updates */
> > +
> > +void typec_set_data_role(struct typec_port *port, enum typec_data_role role)
> > +{
> > + port->data_role = role;
> > + sysfs_notify(&port->dev.kobj, NULL, "current_data_role");
> > +}
> > +EXPORT_SYMBOL(typec_set_data_role);
>
> EXPORT_SYMBOL_GPL() everywhere please, don't mix and match for no good
> reason.
I'll fix these.
> > +static void typec_init_modes(struct typec_altmode *alt, int is_port)
> > +{
> > + struct typec_mode *mode = alt->modes;
> > + int i;
> > +
> > + for (i = 0; i < alt->n_modes; i++, mode++) {
> > + mode->alt_mode = alt;
> > + mode->index = i;
> > + sprintf(mode->group_name, "mode%d", i);
> > +
> > + sysfs_attr_init(&mode->vdo_attr.attr);
> > + mode->vdo_attr.attr.name = "vdo";
> > + mode->vdo_attr.attr.mode = S_IRUGO;
> > + mode->vdo_attr.show = typec_altmode_vdo_show;
> > +
> > + sysfs_attr_init(&mode->desc_attr.attr);
> > + mode->desc_attr.attr.name = "description";
> > + mode->desc_attr.attr.mode = S_IRUGO;
> > + mode->desc_attr.show = typec_altmode_desc_show;
> > +
> > + sysfs_attr_init(&mode->active_attr.attr);
> > + mode->active_attr.attr.name = "active";
> > + mode->active_attr.attr.mode = S_IWUSR | S_IRUGO;
> > + mode->active_attr.show = typec_altmode_active_show;
> > + mode->active_attr.store = typec_altmode_active_store;
> > +
> > + mode->attrs[0] = &mode->vdo_attr.attr;
> > + mode->attrs[1] = &mode->desc_attr.attr;
> > + mode->attrs[2] = &mode->active_attr.attr;
> > +
> > + /* With ports, list the roles that the mode is supported with */
> > + if (is_port) {
> > + sysfs_attr_init(&mode->roles_attr.attr);
> > + mode->roles_attr.attr.name = "supported_roles";
> > + mode->roles_attr.attr.mode = S_IRUGO;
> > + mode->roles_attr.show = typec_altmode_roles_show;
> > +
> > + mode->attrs[3] = &mode->roles_attr.attr;
> > + }
> > +
> > + mode->group.attrs = mode->attrs;
> > + mode->group.name = mode->group_name;
> > +
> > + alt->mode_groups[i] = &mode->group;
> > + }
> > +}
>
> Ugh, dynamic attributes on the fly? why? Why not just use the callback
> to determine if the attribute should be shown or not? Much simpler and
> you don't have to clean up after yourself. Are you sure you cleaned up
> properly from these?
I'll change that.
> > +static ssize_t current_power_role_show(struct device *dev,
> > + struct device_attribute *attr, char *buf)
> > +{
> > + struct typec_port *port = to_typec_port(dev);
> > +
> > + return sprintf(buf, "%s\n", typec_roles[port->pwr_role]);
> > +}
> > +static DEVICE_ATTR_RW(current_power_role);
>
> Nice, see, you use the macros here, be consistant please...
>
> > +static const struct attribute_group typec_group = {
> > + .attrs = typec_attrs,
> > +};
> > +
> > +static const struct attribute_group *typec_groups[] = {
> > + &typec_group,
> > + NULL,
> > +};
>
> ATTRIBUTE_GROUP() please.
>
> > +static int __init typec_init(void)
> > +{
> > + return class_register(&typec_class);
> > +}
> > +subsys_initcall(typec_init);
> > +
> > +static void __exit typec_exit(void)
> > +{
> > + class_unregister(&typec_class);
>
> You forgot to clean up your idr :(
Sorry, what idr? The port ids get removed in typec_release().
Thanks,
--
heikki
On Mon, Nov 14, 2016 at 02:32:35PM +0200, Heikki Krogerus wrote:
> > > +static void __exit typec_exit(void)
> > > +{
> > > + class_unregister(&typec_class);
> >
> > You forgot to clean up your idr :(
>
> Sorry, what idr? The port ids get removed in typec_release().
You have a static idr structure in the driver, right? You have to clean
it up when your code is going away so that it will free any memory it
had allocated with a call to idr_destroy() on module exit.
thanks,
greg k-h
On 11/14/2016 04:32 AM, Heikki Krogerus wrote:
> Hi Greg,
>
> On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
>> On Mon, Sep 19, 2016 at 02:16:56PM +0300, Heikki Krogerus wrote:
>>> The purpose of USB Type-C connector class is to provide
>>> unified interface for the user space to get the status and
>>> basic information about USB Type-C connectors on a system,
>>> control over data role swapping, and when the port supports
>>> USB Power Delivery, also control over power role swapping
>>> and Alternate Modes.
>>>
>>> Reviewed-by: Guenter Roeck <[email protected]>
>>> Tested-by: Guenter Roeck <[email protected]>
>>> Signed-off-by: Heikki Krogerus <[email protected]>
[ ... ]
>>> +
>>> +int typec_connect(struct typec_port *port, struct typec_connection *con)
>>> +{
>>> + int ret;
>>> +
>>> + if (!con->partner && !con->cable)
>>> + return -EINVAL;
>>> +
>>> + port->connected = 1;
>>> + port->data_role = con->data_role;
>>> + port->pwr_role = con->pwr_role;
>>> + port->vconn_role = con->vconn_role;
>>> + port->pwr_opmode = con->pwr_opmode;
>>> +
>>> + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
>>
>> This worries me. Who is listening for it? What will you do with it?
>> Shouldn't you just poll on an attribute file instead?
>
> Oliver! Did you need this or can we remove it?
>
I'll also have to make sure that the Android folks don't use it.
Guenter
On Mon, Nov 14, 2016 at 03:11:23PM +0100, Greg KH wrote:
> On Mon, Nov 14, 2016 at 02:32:35PM +0200, Heikki Krogerus wrote:
> > > > +static void __exit typec_exit(void)
> > > > +{
> > > > + class_unregister(&typec_class);
> > >
> > > You forgot to clean up your idr :(
> >
> > Sorry, what idr? The port ids get removed in typec_release().
>
> You have a static idr structure in the driver, right? You have to clean
> it up when your code is going away so that it will free any memory it
> had allocated with a call to idr_destroy() on module exit.
Ok.
Regarding the DEVICE_ATTR* macros. So I have attributes with same
names for different device types. I may be able to identify the device
types and deal with the correct attribute based on that, but for
example the attribute "active" with alternate modes is writable, but
with cables it's not. How do I handle those?
Thanks,
--
heikki
On Mon, Nov 14, 2016 at 04:39:10PM +0200, Heikki Krogerus wrote:
> On Mon, Nov 14, 2016 at 03:11:23PM +0100, Greg KH wrote:
> > On Mon, Nov 14, 2016 at 02:32:35PM +0200, Heikki Krogerus wrote:
> > > > > +static void __exit typec_exit(void)
> > > > > +{
> > > > > + class_unregister(&typec_class);
> > > >
> > > > You forgot to clean up your idr :(
> > >
> > > Sorry, what idr? The port ids get removed in typec_release().
> >
> > You have a static idr structure in the driver, right? You have to clean
> > it up when your code is going away so that it will free any memory it
> > had allocated with a call to idr_destroy() on module exit.
>
> Ok.
>
> Regarding the DEVICE_ATTR* macros. So I have attributes with same
> names for different device types. I may be able to identify the device
> types and deal with the correct attribute based on that, but for
> example the attribute "active" with alternate modes is writable, but
> with cables it's not. How do I handle those?
The attribute init callback should let you handle this, right? You'll
have to add it for your dynamic attributes.
thanks,
greg k-h
On Mon, Nov 14, 2016 at 02:32:35PM +0200, Heikki Krogerus wrote:
> Hi Greg,
>
> On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
> > On Mon, Sep 19, 2016 at 02:16:56PM +0300, Heikki Krogerus wrote:
> > > The purpose of USB Type-C connector class is to provide
> > > unified interface for the user space to get the status and
> > > basic information about USB Type-C connectors on a system,
> > > control over data role swapping, and when the port supports
> > > USB Power Delivery, also control over power role swapping
> > > and Alternate Modes.
> > >
> > > Reviewed-by: Guenter Roeck <[email protected]>
> > > Tested-by: Guenter Roeck <[email protected]>
> > > Signed-off-by: Heikki Krogerus <[email protected]>
> > > ---
> > > Documentation/ABI/testing/sysfs-class-typec | 218 ++++++
> > > Documentation/usb/typec.txt | 103 +++
> > > MAINTAINERS | 9 +
> > > drivers/usb/Kconfig | 2 +
> > > drivers/usb/Makefile | 2 +
> > > drivers/usb/typec/Kconfig | 7 +
> > > drivers/usb/typec/Makefile | 1 +
> > > drivers/usb/typec/typec.c | 1075 +++++++++++++++++++++++++++
> > > include/linux/usb/typec.h | 252 +++++++
> > > 9 files changed, 1669 insertions(+)
> > > create mode 100644 Documentation/ABI/testing/sysfs-class-typec
> > > create mode 100644 Documentation/usb/typec.txt
> > > create mode 100644 drivers/usb/typec/Kconfig
> > > create mode 100644 drivers/usb/typec/Makefile
> > > create mode 100644 drivers/usb/typec/typec.c
> > > create mode 100644 include/linux/usb/typec.h
> >
[ ... ]
> > > +
> > > +int typec_connect(struct typec_port *port, struct typec_connection *con)
> > > +{
> > > + int ret;
> > > +
> > > + if (!con->partner && !con->cable)
> > > + return -EINVAL;
> > > +
> > > + port->connected = 1;
> > > + port->data_role = con->data_role;
> > > + port->pwr_role = con->pwr_role;
> > > + port->vconn_role = con->vconn_role;
> > > + port->pwr_opmode = con->pwr_opmode;
> > > +
> > > + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
> >
> > This worries me. Who is listening for it? What will you do with it?
> > Shouldn't you just poll on an attribute file instead?
>
> Oliver! Did you need this or can we remove it?
>
> I remember I removed the "connected" attribute because you did not see
> any use for it at one point. I don't remember the reason exactly why?
>
The Android team tells me that they are currently using the udev events
to track port role changes, and to detect presence of port partner.
Also, there are plans to track changes on usbc*cable to differentiate
between cable attach vs. device being attached on the remote end.
What is the problem with using kobject_uevent() and thus presumably
udev events ?
Thanks,
Guenter
On Mon, Nov 14, 2016 at 12:46:50PM -0800, Guenter Roeck wrote:
> On Mon, Nov 14, 2016 at 02:32:35PM +0200, Heikki Krogerus wrote:
> > Hi Greg,
> >
> > On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
> > > On Mon, Sep 19, 2016 at 02:16:56PM +0300, Heikki Krogerus wrote:
> > > > The purpose of USB Type-C connector class is to provide
> > > > unified interface for the user space to get the status and
> > > > basic information about USB Type-C connectors on a system,
> > > > control over data role swapping, and when the port supports
> > > > USB Power Delivery, also control over power role swapping
> > > > and Alternate Modes.
> > > >
> > > > Reviewed-by: Guenter Roeck <[email protected]>
> > > > Tested-by: Guenter Roeck <[email protected]>
> > > > Signed-off-by: Heikki Krogerus <[email protected]>
> > > > ---
> > > > Documentation/ABI/testing/sysfs-class-typec | 218 ++++++
> > > > Documentation/usb/typec.txt | 103 +++
> > > > MAINTAINERS | 9 +
> > > > drivers/usb/Kconfig | 2 +
> > > > drivers/usb/Makefile | 2 +
> > > > drivers/usb/typec/Kconfig | 7 +
> > > > drivers/usb/typec/Makefile | 1 +
> > > > drivers/usb/typec/typec.c | 1075 +++++++++++++++++++++++++++
> > > > include/linux/usb/typec.h | 252 +++++++
> > > > 9 files changed, 1669 insertions(+)
> > > > create mode 100644 Documentation/ABI/testing/sysfs-class-typec
> > > > create mode 100644 Documentation/usb/typec.txt
> > > > create mode 100644 drivers/usb/typec/Kconfig
> > > > create mode 100644 drivers/usb/typec/Makefile
> > > > create mode 100644 drivers/usb/typec/typec.c
> > > > create mode 100644 include/linux/usb/typec.h
> > >
> [ ... ]
>
> > > > +
> > > > +int typec_connect(struct typec_port *port, struct typec_connection *con)
> > > > +{
> > > > + int ret;
> > > > +
> > > > + if (!con->partner && !con->cable)
> > > > + return -EINVAL;
> > > > +
> > > > + port->connected = 1;
> > > > + port->data_role = con->data_role;
> > > > + port->pwr_role = con->pwr_role;
> > > > + port->vconn_role = con->vconn_role;
> > > > + port->pwr_opmode = con->pwr_opmode;
> > > > +
> > > > + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
> > >
> > > This worries me. Who is listening for it? What will you do with it?
> > > Shouldn't you just poll on an attribute file instead?
> >
> > Oliver! Did you need this or can we remove it?
> >
> > I remember I removed the "connected" attribute because you did not see
> > any use for it at one point. I don't remember the reason exactly why?
> >
>
> The Android team tells me that they are currently using the udev events
> to track port role changes, and to detect presence of port partner.
>
> Also, there are plans to track changes on usbc*cable to differentiate
> between cable attach vs. device being attached on the remote end.
>
> What is the problem with using kobject_uevent() and thus presumably
> udev events ?
It's not a "normal" thing to do and is pretty "heavy" to do. What does
userspace do with that change event? Does it read specific attributes?
What causes the event to happen in the kernel, is it really just a
change in the specific object, or do new ones get added/removed?
In short, document the heck out of this please so people know how to use
it, and what is happening when the event happens.
thanks,
greg k-h
On 11/14/2016 11:07 PM, Greg KH wrote:
> On Mon, Nov 14, 2016 at 12:46:50PM -0800, Guenter Roeck wrote:
>> On Mon, Nov 14, 2016 at 02:32:35PM +0200, Heikki Krogerus wrote:
>>> Hi Greg,
>>>
>>> On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
>>>> On Mon, Sep 19, 2016 at 02:16:56PM +0300, Heikki Krogerus wrote:
>>>>> The purpose of USB Type-C connector class is to provide
>>>>> unified interface for the user space to get the status and
>>>>> basic information about USB Type-C connectors on a system,
>>>>> control over data role swapping, and when the port supports
>>>>> USB Power Delivery, also control over power role swapping
>>>>> and Alternate Modes.
>>>>>
>>>>> Reviewed-by: Guenter Roeck <[email protected]>
>>>>> Tested-by: Guenter Roeck <[email protected]>
>>>>> Signed-off-by: Heikki Krogerus <[email protected]>
>>>>> ---
>>>>> Documentation/ABI/testing/sysfs-class-typec | 218 ++++++
>>>>> Documentation/usb/typec.txt | 103 +++
>>>>> MAINTAINERS | 9 +
>>>>> drivers/usb/Kconfig | 2 +
>>>>> drivers/usb/Makefile | 2 +
>>>>> drivers/usb/typec/Kconfig | 7 +
>>>>> drivers/usb/typec/Makefile | 1 +
>>>>> drivers/usb/typec/typec.c | 1075 +++++++++++++++++++++++++++
>>>>> include/linux/usb/typec.h | 252 +++++++
>>>>> 9 files changed, 1669 insertions(+)
>>>>> create mode 100644 Documentation/ABI/testing/sysfs-class-typec
>>>>> create mode 100644 Documentation/usb/typec.txt
>>>>> create mode 100644 drivers/usb/typec/Kconfig
>>>>> create mode 100644 drivers/usb/typec/Makefile
>>>>> create mode 100644 drivers/usb/typec/typec.c
>>>>> create mode 100644 include/linux/usb/typec.h
>>>>
>> [ ... ]
>>
>>>>> +
>>>>> +int typec_connect(struct typec_port *port, struct typec_connection *con)
>>>>> +{
>>>>> + int ret;
>>>>> +
>>>>> + if (!con->partner && !con->cable)
>>>>> + return -EINVAL;
>>>>> +
>>>>> + port->connected = 1;
>>>>> + port->data_role = con->data_role;
>>>>> + port->pwr_role = con->pwr_role;
>>>>> + port->vconn_role = con->vconn_role;
>>>>> + port->pwr_opmode = con->pwr_opmode;
>>>>> +
>>>>> + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
>>>>
>>>> This worries me. Who is listening for it? What will you do with it?
>>>> Shouldn't you just poll on an attribute file instead?
>>>
>>> Oliver! Did you need this or can we remove it?
>>>
>>> I remember I removed the "connected" attribute because you did not see
>>> any use for it at one point. I don't remember the reason exactly why?
>>>
>>
>> The Android team tells me that they are currently using the udev events
>> to track port role changes, and to detect presence of port partner.
>>
>> Also, there are plans to track changes on usbc*cable to differentiate
>> between cable attach vs. device being attached on the remote end.
>>
>> What is the problem with using kobject_uevent() and thus presumably
>> udev events ?
>
> It's not a "normal" thing to do and is pretty "heavy" to do. What does
> userspace do with that change event? Does it read specific attributes?
> What causes the event to happen in the kernel, is it really just a
> change in the specific object, or do new ones get added/removed?
>
> In short, document the heck out of this please so people know how to use
> it, and what is happening when the event happens.
>
Badhri,
can you clarify which events you are using in detail, and what for ?
Thanks,
Guenter
Hi,
At present I am using the uevent in the userspace to infer
the Presence of a port on the remote end through the
appearance of usbc*-partner.
Userspace uses this info to decide on when to show a USB
notification on the screen and what should be the options
provided in the dialog.
I was assuming that this is not something that would be dropped.
Coding using events was relatively easier to program from userspace ..
Is it possible to use POLL for identifying the appearance of port partner ?
I did not notice sysfs_notify call in typec_connect/typec_disconnect.
It would also be nice to have uevent notifications when the contents
of current_data_role or current_power_role changes.
Is that too costly to have ?
Thanks,
Badhri.
On Tue, Nov 15, 2016 at 1:25 AM, Guenter Roeck <[email protected]> wrote:
> On 11/14/2016 11:07 PM, Greg KH wrote:
>>
>> On Mon, Nov 14, 2016 at 12:46:50PM -0800, Guenter Roeck wrote:
>>>
>>> On Mon, Nov 14, 2016 at 02:32:35PM +0200, Heikki Krogerus wrote:
>>>>
>>>> Hi Greg,
>>>>
>>>> On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
>>>>>
>>>>> On Mon, Sep 19, 2016 at 02:16:56PM +0300, Heikki Krogerus wrote:
>>>>>>
>>>>>> The purpose of USB Type-C connector class is to provide
>>>>>> unified interface for the user space to get the status and
>>>>>> basic information about USB Type-C connectors on a system,
>>>>>> control over data role swapping, and when the port supports
>>>>>> USB Power Delivery, also control over power role swapping
>>>>>> and Alternate Modes.
>>>>>>
>>>>>> Reviewed-by: Guenter Roeck <[email protected]>
>>>>>> Tested-by: Guenter Roeck <[email protected]>
>>>>>> Signed-off-by: Heikki Krogerus <[email protected]>
>>>>>> ---
>>>>>> Documentation/ABI/testing/sysfs-class-typec | 218 ++++++
>>>>>> Documentation/usb/typec.txt | 103 +++
>>>>>> MAINTAINERS | 9 +
>>>>>> drivers/usb/Kconfig | 2 +
>>>>>> drivers/usb/Makefile | 2 +
>>>>>> drivers/usb/typec/Kconfig | 7 +
>>>>>> drivers/usb/typec/Makefile | 1 +
>>>>>> drivers/usb/typec/typec.c | 1075
>>>>>> +++++++++++++++++++++++++++
>>>>>> include/linux/usb/typec.h | 252 +++++++
>>>>>> 9 files changed, 1669 insertions(+)
>>>>>> create mode 100644 Documentation/ABI/testing/sysfs-class-typec
>>>>>> create mode 100644 Documentation/usb/typec.txt
>>>>>> create mode 100644 drivers/usb/typec/Kconfig
>>>>>> create mode 100644 drivers/usb/typec/Makefile
>>>>>> create mode 100644 drivers/usb/typec/typec.c
>>>>>> create mode 100644 include/linux/usb/typec.h
>>>>>
>>>>>
>>> [ ... ]
>>>
>>>>>> +
>>>>>> +int typec_connect(struct typec_port *port, struct typec_connection
>>>>>> *con)
>>>>>> +{
>>>>>> + int ret;
>>>>>> +
>>>>>> + if (!con->partner && !con->cable)
>>>>>> + return -EINVAL;
>>>>>> +
>>>>>> + port->connected = 1;
>>>>>> + port->data_role = con->data_role;
>>>>>> + port->pwr_role = con->pwr_role;
>>>>>> + port->vconn_role = con->vconn_role;
>>>>>> + port->pwr_opmode = con->pwr_opmode;
>>>>>> +
>>>>>> + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
>>>>>
>>>>>
>>>>> This worries me. Who is listening for it? What will you do with it?
>>>>> Shouldn't you just poll on an attribute file instead?
>>>>
>>>>
>>>> Oliver! Did you need this or can we remove it?
>>>>
>>>> I remember I removed the "connected" attribute because you did not see
>>>> any use for it at one point. I don't remember the reason exactly why?
>>>>
>>>
>>> The Android team tells me that they are currently using the udev events
>>> to track port role changes, and to detect presence of port partner.
>>>
>>> Also, there are plans to track changes on usbc*cable to differentiate
>>> between cable attach vs. device being attached on the remote end.
>>>
>>> What is the problem with using kobject_uevent() and thus presumably
>>> udev events ?
>>
>>
>> It's not a "normal" thing to do and is pretty "heavy" to do. What does
>> userspace do with that change event? Does it read specific attributes?
>> What causes the event to happen in the kernel, is it really just a
>> change in the specific object, or do new ones get added/removed?
>>
>> In short, document the heck out of this please so people know how to use
>> it, and what is happening when the event happens.
>>
>
> Badhri,
>
> can you clarify which events you are using in detail, and what for ?
>
> Thanks,
> Guenter
>
On Mon, 2016-11-14 at 06:34 -0800, Guenter Roeck wrote:
> >>> +int typec_connect(struct typec_port *port, struct
> typec_connection *con)
> >>> +{
> >>> + int ret;
> >>> +
> >>> + if (!con->partner && !con->cable)
> >>> + return -EINVAL;
> >>> +
> >>> + port->connected = 1;
> >>> + port->data_role = con->data_role;
> >>> + port->pwr_role = con->pwr_role;
> >>> + port->vconn_role = con->vconn_role;
> >>> + port->pwr_opmode = con->pwr_opmode;
> >>> +
> >>> + kobject_uevent(&port->dev.kobj, KOBJ_CHANGE);
> >>
> >> This worries me. Who is listening for it? What will you do with
> it?
> >> Shouldn't you just poll on an attribute file instead?
> >
> > Oliver! Did you need this or can we remove it?
> >
>
> I'll also have to make sure that the Android folks don't use it.
How then do we notify user space? poll()? Yet another demon.
I do not specifically need this, but I note that uevents are the
general tool we use to notify user space of that kind of events.
Regards
Oliver
On Tue, Nov 15, 2016 at 04:19:10PM -0800, Badhri Jagan Sridharan wrote:
> Hi,
>
> At present I am using the uevent in the userspace to infer
> the Presence of a port on the remote end through the
> appearance of usbc*-partner.
>
> Userspace uses this info to decide on when to show a USB
> notification on the screen and what should be the options
> provided in the dialog.
>
> I was assuming that this is not something that would be dropped.
>
> Coding using events was relatively easier to program from userspace ..
>
> Is it possible to use POLL for identifying the appearance of port partner ?
> I did not notice sysfs_notify call in typec_connect/typec_disconnect.
>
> It would also be nice to have uevent notifications when the contents
> of current_data_role or current_power_role changes.
>
> Is that too costly to have ?
Greg, could you give your opinion. In this case we do have attribute
files that the user space can poll. Data role is the USB data role, so
host or device, and it can change for example if the partner executes
a swap. The same can happen with the power role.
Thanks,
--
heikki
On Wed, 2016-11-16 at 11:30 +0200, Heikki Krogerus wrote:
> On Tue, Nov 15, 2016 at 04:19:10PM -0800, Badhri Jagan Sridharan wrote:
> > Is that too costly to have ?
>
> Greg, could you give your opinion. In this case we do have attribute
> files that the user space can poll. Data role is the USB data role, so
> host or device, and it can change for example if the partner executes
> a swap. The same can happen with the power role.
IMHO the uevent is cheaper. User space cannot just poll without further
infrastructure. A task needs to run to poll. A uevent can be handled
through established infrastructure.
Sure from a kernel level it is the heavier gun, but I think this is
the wrong angle of looking at this issue.
Regards
Oliver
On Wed, Nov 16, 2016 at 11:30:35AM +0200, Heikki Krogerus wrote:
> On Tue, Nov 15, 2016 at 04:19:10PM -0800, Badhri Jagan Sridharan wrote:
> > Hi,
> >
> > At present I am using the uevent in the userspace to infer
> > the Presence of a port on the remote end through the
> > appearance of usbc*-partner.
> >
> > Userspace uses this info to decide on when to show a USB
> > notification on the screen and what should be the options
> > provided in the dialog.
> >
> > I was assuming that this is not something that would be dropped.
> >
> > Coding using events was relatively easier to program from userspace ..
> >
> > Is it possible to use POLL for identifying the appearance of port partner ?
> > I did not notice sysfs_notify call in typec_connect/typec_disconnect.
> >
> > It would also be nice to have uevent notifications when the contents
> > of current_data_role or current_power_role changes.
> >
> > Is that too costly to have ?
>
> Greg, could you give your opinion. In this case we do have attribute
> files that the user space can poll. Data role is the USB data role, so
> host or device, and it can change for example if the partner executes
> a swap. The same can happen with the power role.
So the same 'struct device' switches roles and attribute files are
updated that need to be re-read? If so, yes KOBJ_CHANGE is correct, if
a struct device is added/removed for this, then no, it doesn't make
sense.
Again, document this to describe what is happening and it might be more
obvious to people and so these questions would not come up :)
thanks,
greg k-h
On Wed, Nov 16, 2016 at 10:49:49AM +0100, Greg KH wrote:
> On Wed, Nov 16, 2016 at 11:30:35AM +0200, Heikki Krogerus wrote:
> > On Tue, Nov 15, 2016 at 04:19:10PM -0800, Badhri Jagan Sridharan wrote:
> > > Hi,
> > >
> > > At present I am using the uevent in the userspace to infer
> > > the Presence of a port on the remote end through the
> > > appearance of usbc*-partner.
> > >
> > > Userspace uses this info to decide on when to show a USB
> > > notification on the screen and what should be the options
> > > provided in the dialog.
> > >
> > > I was assuming that this is not something that would be dropped.
> > >
> > > Coding using events was relatively easier to program from userspace ..
> > >
> > > Is it possible to use POLL for identifying the appearance of port partner ?
> > > I did not notice sysfs_notify call in typec_connect/typec_disconnect.
> > >
> > > It would also be nice to have uevent notifications when the contents
> > > of current_data_role or current_power_role changes.
> > >
> > > Is that too costly to have ?
> >
> > Greg, could you give your opinion. In this case we do have attribute
> > files that the user space can poll. Data role is the USB data role, so
> > host or device, and it can change for example if the partner executes
> > a swap. The same can happen with the power role.
>
> So the same 'struct device' switches roles and attribute files are
> updated that need to be re-read? If so, yes KOBJ_CHANGE is correct, if
> a struct device is added/removed for this, then no, it doesn't make
> sense.
OK, I'll add KOBJ_CHANGE for those.
So is it OK to everybody if I remove the KOBJ_CHANGE in
typec_connect()? We will see uevent KOBJ_ADD since the partner (or
cable) is added in any case. Badhri, Oliver?
Thanks,
--
heikki
On Wed, 2016-11-16 at 13:09 +0200, Heikki Krogerus wrote:
> OK, I'll add KOBJ_CHANGE for those.
>
> So is it OK to everybody if I remove the KOBJ_CHANGE in
> typec_connect()? We will see uevent KOBJ_ADD since the partner (or
> cable) is added in any case. Badhri, Oliver?
OK by me.
Regards
Oliver
> IMHO the uevent is cheaper. User space cannot just poll without further
> infrastructure. A task needs to run to poll. A uevent can be handled
> through established infrastructure.
Thanks Oliver for stating this. This is exactly what I was facing.
> OK, I'll add KOBJ_CHANGE for those.
>
> So is it OK to everybody if I remove the KOBJ_CHANGE in
> typec_connect()? We will see uevent KOBJ_ADD since the partner (or
> cable) is added in any case. Badhri, Oliver?
Yes Heikki.. That's OK for me as well.
Just to get my understanding right. You are planning to add
KOBJ_CHANGE uevents when current_power_role or
current_data_role changes and KOBJ_ADD when new port-partner
or the cable is attached. Is that right ?
Thanks,
Badhri.
On Wed, Nov 16, 2016 at 3:27 AM, Oliver Neukum <[email protected]> wrote:
> On Wed, 2016-11-16 at 13:09 +0200, Heikki Krogerus wrote:
>
>> OK, I'll add KOBJ_CHANGE for those.
>>
>> So is it OK to everybody if I remove the KOBJ_CHANGE in
>> typec_connect()? We will see uevent KOBJ_ADD since the partner (or
>> cable) is added in any case. Badhri, Oliver?
>
> OK by me.
>
> Regards
> Oliver
>
>
On Wed, Nov 16, 2016 at 06:30:23AM -0800, Badhri Jagan Sridharan wrote:
> > IMHO the uevent is cheaper. User space cannot just poll without further
> > infrastructure. A task needs to run to poll. A uevent can be handled
> > through established infrastructure.
>
> Thanks Oliver for stating this. This is exactly what I was facing.
>
> > OK, I'll add KOBJ_CHANGE for those.
> >
> > So is it OK to everybody if I remove the KOBJ_CHANGE in
> > typec_connect()? We will see uevent KOBJ_ADD since the partner (or
> > cable) is added in any case. Badhri, Oliver?
>
> Yes Heikki.. That's OK for me as well.
> Just to get my understanding right. You are planning to add
> KOBJ_CHANGE uevents when current_power_role or
> current_data_role changes and KOBJ_ADD when new port-partner
> or the cable is attached. Is that right ?
Yes, though I don't KOBJ_ADD separately with the partners and cables.
That uevent is sent when the device for them is registered, so it's
already there.
Br,
--
heikki
Hi Greg,
On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
> > +static int sysfs_strmatch(const char * const *array, size_t n, const char *str)
> > +{
> > + const char *item;
> > + int index;
> > +
> > + for (index = 0; index < n; index++) {
> > + item = array[index];
> > + if (!item)
> > + break;
> > + if (sysfs_streq(item, str))
> > + return index;
> > + }
> > +
> > + return -EINVAL;
> > +}
>
> should we make this a core sysfs function?
Last question before I send v11. Is the following (the helper) OK?
diff --git a/include/linux/string.h b/include/linux/string.h
index 26b6f6a..5606810 100644
--- a/include/linux/string.h
+++ b/include/linux/string.h
@@ -135,6 +135,16 @@ static inline int strtobool(const char *s, bool *res)
}
int match_string(const char * const *array, size_t n, const char *string);
+int __sysfs_strmatch(const char * const *array, size_t n, const char *string);
+
+/**
+ * sysfs_strmatch - matches given string in an array
+ * @a: array of strings
+ * @s: string to match with
+ *
+ * Helper for __sysfs_strmatch(). Calculates the size of @a automatically.
+ */
+#define sysfs_strmatch(a, s) __sysfs_strmatch(a, ARRAY_SIZE(a), s)
#ifdef CONFIG_BINARY_PRINTF
int vbin_printf(u32 *bin_buf, size_t size, const char *fmt, va_list args);
diff --git a/lib/string.c b/lib/string.c
index ed83562..a4fe035 100644
--- a/lib/string.c
+++ b/lib/string.c
@@ -656,6 +656,32 @@ int match_string(const char * const *array, size_t n, const char *string)
}
EXPORT_SYMBOL(match_string);
+/**
+ * __sysfs_strmatch - matches given string in an array
+ * @array: array of strings
+ * @n: number of strings in the array or -1 for NULL terminated arrays
+ * @str: string to match with
+ *
+ * Returns index of @str in the @array or -EINVAL, just like match_string().
+ * Uses sysfs_streq() instead of strcmp for matching.
+ */
+int __sysfs_strmatch(const char * const *array, size_t n, const char *str)
+{
+ const char *item;
+ int index;
+
+ for (index = 0; index < n; index++) {
+ item = array[index];
+ if (!item)
+ break;
+ if (!sysfs_streq(item, str))
+ return index;
+ }
+
+ return -EINVAL;
+}
+EXPORT_SYMBOL(__sysfs_strmatch);
+
#ifndef __HAVE_ARCH_MEMSET
/**
* memset - Fill a region of memory with the given value
--
heikki
> Yes, though I don't KOBJ_ADD separately with the partners and cables.
> That uevent is sent when the device for them is registered, so it's
> already there
Makes sense. Thanks Heikki.
Regards,
Badhri
On Wed, Nov 16, 2016 at 7:20 AM, Heikki Krogerus
<[email protected]> wrote:
> Hi Greg,
>
> On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
>> > +static int sysfs_strmatch(const char * const *array, size_t n, const char *str)
>> > +{
>> > + const char *item;
>> > + int index;
>> > +
>> > + for (index = 0; index < n; index++) {
>> > + item = array[index];
>> > + if (!item)
>> > + break;
>> > + if (sysfs_streq(item, str))
>> > + return index;
>> > + }
>> > +
>> > + return -EINVAL;
>> > +}
>>
>> should we make this a core sysfs function?
>
> Last question before I send v11. Is the following (the helper) OK?
>
>
> diff --git a/include/linux/string.h b/include/linux/string.h
> index 26b6f6a..5606810 100644
> --- a/include/linux/string.h
> +++ b/include/linux/string.h
> @@ -135,6 +135,16 @@ static inline int strtobool(const char *s, bool *res)
> }
>
> int match_string(const char * const *array, size_t n, const char *string);
> +int __sysfs_strmatch(const char * const *array, size_t n, const char *string);
> +
> +/**
> + * sysfs_strmatch - matches given string in an array
> + * @a: array of strings
> + * @s: string to match with
> + *
> + * Helper for __sysfs_strmatch(). Calculates the size of @a automatically.
> + */
> +#define sysfs_strmatch(a, s) __sysfs_strmatch(a, ARRAY_SIZE(a), s)
>
> #ifdef CONFIG_BINARY_PRINTF
> int vbin_printf(u32 *bin_buf, size_t size, const char *fmt, va_list args);
> diff --git a/lib/string.c b/lib/string.c
> index ed83562..a4fe035 100644
> --- a/lib/string.c
> +++ b/lib/string.c
> @@ -656,6 +656,32 @@ int match_string(const char * const *array, size_t n, const char *string)
> }
> EXPORT_SYMBOL(match_string);
>
> +/**
> + * __sysfs_strmatch - matches given string in an array
> + * @array: array of strings
> + * @n: number of strings in the array or -1 for NULL terminated arrays
> + * @str: string to match with
> + *
> + * Returns index of @str in the @array or -EINVAL, just like match_string().
> + * Uses sysfs_streq() instead of strcmp for matching.
> + */
> +int __sysfs_strmatch(const char * const *array, size_t n, const char *str)
> +{
> + const char *item;
> + int index;
> +
> + for (index = 0; index < n; index++) {
> + item = array[index];
> + if (!item)
> + break;
> + if (!sysfs_streq(item, str))
> + return index;
> + }
> +
> + return -EINVAL;
> +}
> +EXPORT_SYMBOL(__sysfs_strmatch);
> +
> #ifndef __HAVE_ARCH_MEMSET
> /**
> * memset - Fill a region of memory with the given value
>
>
> --
> heikki
> --
> To unsubscribe from this list: send the line "unsubscribe linux-usb" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, Nov 16, 2016 at 05:20:24PM +0200, Heikki Krogerus wrote:
> Hi Greg,
>
> On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
> > > +static int sysfs_strmatch(const char * const *array, size_t n, const char *str)
> > > +{
> > > + const char *item;
> > > + int index;
> > > +
> > > + for (index = 0; index < n; index++) {
> > > + item = array[index];
> > > + if (!item)
> > > + break;
> > > + if (sysfs_streq(item, str))
> > > + return index;
> > > + }
> > > +
> > > + return -EINVAL;
> > > +}
> >
> > should we make this a core sysfs function?
>
> Last question before I send v11. Is the following (the helper) OK?
>
>
> diff --git a/include/linux/string.h b/include/linux/string.h
> index 26b6f6a..5606810 100644
> --- a/include/linux/string.h
> +++ b/include/linux/string.h
> @@ -135,6 +135,16 @@ static inline int strtobool(const char *s, bool *res)
> }
>
> int match_string(const char * const *array, size_t n, const char *string);
> +int __sysfs_strmatch(const char * const *array, size_t n, const char *string);
> +
> +/**
> + * sysfs_strmatch - matches given string in an array
> + * @a: array of strings
> + * @s: string to match with
> + *
> + * Helper for __sysfs_strmatch(). Calculates the size of @a automatically.
> + */
> +#define sysfs_strmatch(a, s) __sysfs_strmatch(a, ARRAY_SIZE(a), s)
People will bikeshed the name. Why not just use sysfs_match_string() as
this does the same as match_string, but calls sysfs_string instead of
strcmp().
thanks,
greg k-h
On Wed, Nov 16, 2016 at 04:31:07PM +0100, Greg KH wrote:
> On Wed, Nov 16, 2016 at 05:20:24PM +0200, Heikki Krogerus wrote:
> > Hi Greg,
> >
> > On Mon, Nov 14, 2016 at 10:51:48AM +0100, Greg KH wrote:
> > > > +static int sysfs_strmatch(const char * const *array, size_t n, const char *str)
> > > > +{
> > > > + const char *item;
> > > > + int index;
> > > > +
> > > > + for (index = 0; index < n; index++) {
> > > > + item = array[index];
> > > > + if (!item)
> > > > + break;
> > > > + if (sysfs_streq(item, str))
> > > > + return index;
> > > > + }
> > > > +
> > > > + return -EINVAL;
> > > > +}
> > >
> > > should we make this a core sysfs function?
> >
> > Last question before I send v11. Is the following (the helper) OK?
> >
> >
> > diff --git a/include/linux/string.h b/include/linux/string.h
> > index 26b6f6a..5606810 100644
> > --- a/include/linux/string.h
> > +++ b/include/linux/string.h
> > @@ -135,6 +135,16 @@ static inline int strtobool(const char *s, bool *res)
> > }
> >
> > int match_string(const char * const *array, size_t n, const char *string);
> > +int __sysfs_strmatch(const char * const *array, size_t n, const char *string);
> > +
> > +/**
> > + * sysfs_strmatch - matches given string in an array
> > + * @a: array of strings
> > + * @s: string to match with
> > + *
> > + * Helper for __sysfs_strmatch(). Calculates the size of @a automatically.
> > + */
> > +#define sysfs_strmatch(a, s) __sysfs_strmatch(a, ARRAY_SIZE(a), s)
>
> People will bikeshed the name. Why not just use sysfs_match_string() as
> this does the same as match_string, but calls sysfs_string instead of
> strcmp().
Makes sense. I'll change the name.
Thanks,
--
heikki