With the addition of more link speed mode to the netdev trigger, it was
pointed out that there may be a problem with bloating the attribute list
with modes that won't ever be supported by the trigger as the attached
device name doesn't support them.
To clear and address this problem, change the logic where these
additional trigger modes are added.
Since the netdev trigger REQUIRE a device name to be set, attach to the
device name change function additional logic to parse the supported link
speed modes using ethtool APIs and add only the supported link speed
modes attribute.
This only apply to the link speed modes and every other mode is still
provided by default.
Signed-off-by: Christian Marangi <[email protected]>
---
drivers/leds/trigger/ledtrig-netdev.c | 56 +++++++++++++++++++++++----
1 file changed, 49 insertions(+), 7 deletions(-)
diff --git a/drivers/leds/trigger/ledtrig-netdev.c b/drivers/leds/trigger/ledtrig-netdev.c
index 09e75fd9f2bc..ce84808e231c 100644
--- a/drivers/leds/trigger/ledtrig-netdev.c
+++ b/drivers/leds/trigger/ledtrig-netdev.c
@@ -61,6 +61,8 @@ struct led_netdev_data {
bool hw_control;
};
+static int add_link_speed_attr(struct led_netdev_data *trigger_data);
+
static void set_baseline_state(struct led_netdev_data *trigger_data)
{
int current_brightness;
@@ -262,8 +264,10 @@ static int set_device_name(struct led_netdev_data *trigger_data,
trigger_data->carrier_link_up = false;
trigger_data->link_speed = SPEED_UNKNOWN;
trigger_data->duplex = DUPLEX_UNKNOWN;
- if (trigger_data->net_dev)
+ if (trigger_data->net_dev) {
get_device_state(trigger_data);
+ add_link_speed_attr(trigger_data);
+ }
trigger_data->last_activity = 0;
@@ -396,6 +400,50 @@ DEFINE_NETDEV_TRIGGER(full_duplex, TRIGGER_NETDEV_FULL_DUPLEX);
DEFINE_NETDEV_TRIGGER(tx, TRIGGER_NETDEV_TX);
DEFINE_NETDEV_TRIGGER(rx, TRIGGER_NETDEV_RX);
+static int add_link_speed_attr(struct led_netdev_data *trigger_data)
+{
+ struct led_classdev *led_cdev = trigger_data->led_cdev;
+ struct device *dev = led_cdev->dev;
+ struct ethtool_link_ksettings cmd;
+ int ret;
+
+ /* First remove any entry previously added */
+ device_remove_file(dev, &dev_attr_link_10);
+ device_remove_file(dev, &dev_attr_link_100);
+ device_remove_file(dev, &dev_attr_link_1000);
+ device_remove_file(dev, &dev_attr_link_2500);
+ device_remove_file(dev, &dev_attr_link_5000);
+ device_remove_file(dev, &dev_attr_link_10000);
+
+ ret = __ethtool_get_link_ksettings(trigger_data->net_dev, &cmd);
+ if (ret)
+ return ret;
+
+ /* Add only supported entry */
+ if (test_bit(ETHTOOL_LINK_MODE_10baseT_Half_BIT, cmd.link_modes.supported) ||
+ test_bit(ETHTOOL_LINK_MODE_10baseT_Full_BIT, cmd.link_modes.supported))
+ device_create_file(dev, &dev_attr_link_10);
+
+ if (test_bit(ETHTOOL_LINK_MODE_100baseT_Half_BIT, cmd.link_modes.supported) ||
+ test_bit(ETHTOOL_LINK_MODE_100baseT_Full_BIT, cmd.link_modes.supported))
+ device_create_file(dev, &dev_attr_link_100);
+
+ if (test_bit(ETHTOOL_LINK_MODE_1000baseT_Half_BIT, cmd.link_modes.supported) ||
+ test_bit(ETHTOOL_LINK_MODE_1000baseT_Full_BIT, cmd.link_modes.supported))
+ device_create_file(dev, &dev_attr_link_1000);
+
+ if (test_bit(ETHTOOL_LINK_MODE_2500baseT_Full_BIT, cmd.link_modes.supported))
+ device_create_file(dev, &dev_attr_link_2500);
+
+ if (test_bit(ETHTOOL_LINK_MODE_5000baseT_Full_BIT, cmd.link_modes.supported))
+ device_create_file(dev, &dev_attr_link_5000);
+
+ if (test_bit(ETHTOOL_LINK_MODE_10000baseT_Full_BIT, cmd.link_modes.supported))
+ device_create_file(dev, &dev_attr_link_10000);
+
+ return 0;
+}
+
static ssize_t interval_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
@@ -446,12 +494,6 @@ static DEVICE_ATTR_RO(offloaded);
static struct attribute *netdev_trig_attrs[] = {
&dev_attr_device_name.attr,
&dev_attr_link.attr,
- &dev_attr_link_10.attr,
- &dev_attr_link_100.attr,
- &dev_attr_link_1000.attr,
- &dev_attr_link_2500.attr,
- &dev_attr_link_5000.attr,
- &dev_attr_link_10000.attr,
&dev_attr_full_duplex.attr,
&dev_attr_half_duplex.attr,
&dev_attr_rx.attr,
--
2.40.1
On Sat, Dec 09, 2023 at 04:07:24PM +0100, Christian Marangi wrote:
> With the addition of more link speed mode to the netdev trigger, it was
> pointed out that there may be a problem with bloating the attribute list
> with modes that won't ever be supported by the trigger as the attached
> device name doesn't support them.
>
> To clear and address this problem, change the logic where these
> additional trigger modes are added.
>
> Since the netdev trigger REQUIRE a device name to be set, attach to the
> device name change function additional logic to parse the supported link
> speed modes using ethtool APIs and add only the supported link speed
> modes attribute.
>
> This only apply to the link speed modes and every other mode is still
> provided by default.
Deleting the file and then re-creating it is not so nice when handling
user space. Have you tried this while holding the file open?
Please take a look at the is_visible() method of a sysfs group. If it
returns 0, the file is simply not visible in sysfs, but it still
exists. It is not a problem if a process holds it open when it becomes
invisible.
Andrew
---
pw-bot: cr
On Sat, 9 Dec 2023 16:07:24 +0100
Christian Marangi <[email protected]> wrote:
> With the addition of more link speed mode to the netdev trigger, it was
> pointed out that there may be a problem with bloating the attribute list
> with modes that won't ever be supported by the trigger as the attached
> device name doesn't support them.
>
> To clear and address this problem, change the logic where these
> additional trigger modes are added.
>
> Since the netdev trigger REQUIRE a device name to be set, attach to the
> device name change function additional logic to parse the supported link
> speed modes using ethtool APIs and add only the supported link speed
> modes attribute.
>
> This only apply to the link speed modes and every other mode is still
> provided by default.
>
> Signed-off-by: Christian Marangi <[email protected]>
> ---
> drivers/leds/trigger/ledtrig-netdev.c | 56 +++++++++++++++++++++++----
> 1 file changed, 49 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/leds/trigger/ledtrig-netdev.c b/drivers/leds/trigger/ledtrig-netdev.c
> index 09e75fd9f2bc..ce84808e231c 100644
> --- a/drivers/leds/trigger/ledtrig-netdev.c
> +++ b/drivers/leds/trigger/ledtrig-netdev.c
> @@ -61,6 +61,8 @@ struct led_netdev_data {
> bool hw_control;
> };
>
> +static int add_link_speed_attr(struct led_netdev_data *trigger_data);
> +
> static void set_baseline_state(struct led_netdev_data *trigger_data)
> {
> int current_brightness;
> @@ -262,8 +264,10 @@ static int set_device_name(struct led_netdev_data *trigger_data,
> trigger_data->carrier_link_up = false;
> trigger_data->link_speed = SPEED_UNKNOWN;
> trigger_data->duplex = DUPLEX_UNKNOWN;
> - if (trigger_data->net_dev)
> + if (trigger_data->net_dev) {
> get_device_state(trigger_data);
> + add_link_speed_attr(trigger_data);
> + }
>
> trigger_data->last_activity = 0;
>
> @@ -396,6 +400,50 @@ DEFINE_NETDEV_TRIGGER(full_duplex, TRIGGER_NETDEV_FULL_DUPLEX);
> DEFINE_NETDEV_TRIGGER(tx, TRIGGER_NETDEV_TX);
> DEFINE_NETDEV_TRIGGER(rx, TRIGGER_NETDEV_RX);
>
> +static int add_link_speed_attr(struct led_netdev_data *trigger_data)
> +{
> + struct led_classdev *led_cdev = trigger_data->led_cdev;
> + struct device *dev = led_cdev->dev;
> + struct ethtool_link_ksettings cmd;
> + int ret;
> +
> + /* First remove any entry previously added */
> + device_remove_file(dev, &dev_attr_link_10);
> + device_remove_file(dev, &dev_attr_link_100);
> + device_remove_file(dev, &dev_attr_link_1000);
> + device_remove_file(dev, &dev_attr_link_2500);
> + device_remove_file(dev, &dev_attr_link_5000);
> + device_remove_file(dev, &dev_attr_link_10000);
Noooooo!
> + ret = __ethtool_get_link_ksettings(trigger_data->net_dev, &cmd);
> + if (ret)
> + return ret;
> +
> + /* Add only supported entry */
> + if (test_bit(ETHTOOL_LINK_MODE_10baseT_Half_BIT, cmd.link_modes.supported) ||
> + test_bit(ETHTOOL_LINK_MODE_10baseT_Full_BIT, cmd.link_modes.supported))
> + device_create_file(dev, &dev_attr_link_10);
> +
> + if (test_bit(ETHTOOL_LINK_MODE_100baseT_Half_BIT, cmd.link_modes.supported) ||
> + test_bit(ETHTOOL_LINK_MODE_100baseT_Full_BIT, cmd.link_modes.supported))
> + device_create_file(dev, &dev_attr_link_100);
> +
> + if (test_bit(ETHTOOL_LINK_MODE_1000baseT_Half_BIT, cmd.link_modes.supported) ||
> + test_bit(ETHTOOL_LINK_MODE_1000baseT_Full_BIT, cmd.link_modes.supported))
> + device_create_file(dev, &dev_attr_link_1000);
> +
> + if (test_bit(ETHTOOL_LINK_MODE_2500baseT_Full_BIT, cmd.link_modes.supported))
> + device_create_file(dev, &dev_attr_link_2500);
> +
> + if (test_bit(ETHTOOL_LINK_MODE_5000baseT_Full_BIT, cmd.link_modes.supported))
> + device_create_file(dev, &dev_attr_link_5000);
> +
> + if (test_bit(ETHTOOL_LINK_MODE_10000baseT_Full_BIT, cmd.link_modes.supported))
> + device_create_file(dev, &dev_attr_link_10000);
> +
> + return 0;
> +}
This should be done via the is_visible sysfs attribute_group method.
Also documentation for the link_* files should be changed so that it
says that the files are present only if those speeds are available.
Marek
On Wed, Dec 13, 2023 at 01:05:21PM +0100, Marek Beh?n wrote:
> On Sat, 9 Dec 2023 16:07:24 +0100
> Christian Marangi <[email protected]> wrote:
>
> > With the addition of more link speed mode to the netdev trigger, it was
> > pointed out that there may be a problem with bloating the attribute list
> > with modes that won't ever be supported by the trigger as the attached
> > device name doesn't support them.
> >
> > To clear and address this problem, change the logic where these
> > additional trigger modes are added.
> >
> > Since the netdev trigger REQUIRE a device name to be set, attach to the
> > device name change function additional logic to parse the supported link
> > speed modes using ethtool APIs and add only the supported link speed
> > modes attribute.
> >
> > This only apply to the link speed modes and every other mode is still
> > provided by default.
> >
> > Signed-off-by: Christian Marangi <[email protected]>
> > ---
> > drivers/leds/trigger/ledtrig-netdev.c | 56 +++++++++++++++++++++++----
> > 1 file changed, 49 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/leds/trigger/ledtrig-netdev.c b/drivers/leds/trigger/ledtrig-netdev.c
> > index 09e75fd9f2bc..ce84808e231c 100644
> > --- a/drivers/leds/trigger/ledtrig-netdev.c
> > +++ b/drivers/leds/trigger/ledtrig-netdev.c
> > @@ -61,6 +61,8 @@ struct led_netdev_data {
> > bool hw_control;
> > };
> >
> > +static int add_link_speed_attr(struct led_netdev_data *trigger_data);
> > +
> > static void set_baseline_state(struct led_netdev_data *trigger_data)
> > {
> > int current_brightness;
> > @@ -262,8 +264,10 @@ static int set_device_name(struct led_netdev_data *trigger_data,
> > trigger_data->carrier_link_up = false;
> > trigger_data->link_speed = SPEED_UNKNOWN;
> > trigger_data->duplex = DUPLEX_UNKNOWN;
> > - if (trigger_data->net_dev)
> > + if (trigger_data->net_dev) {
> > get_device_state(trigger_data);
> > + add_link_speed_attr(trigger_data);
> > + }
> >
> > trigger_data->last_activity = 0;
> >
> > @@ -396,6 +400,50 @@ DEFINE_NETDEV_TRIGGER(full_duplex, TRIGGER_NETDEV_FULL_DUPLEX);
> > DEFINE_NETDEV_TRIGGER(tx, TRIGGER_NETDEV_TX);
> > DEFINE_NETDEV_TRIGGER(rx, TRIGGER_NETDEV_RX);
> >
> > +static int add_link_speed_attr(struct led_netdev_data *trigger_data)
> > +{
> > + struct led_classdev *led_cdev = trigger_data->led_cdev;
> > + struct device *dev = led_cdev->dev;
> > + struct ethtool_link_ksettings cmd;
> > + int ret;
> > +
> > + /* First remove any entry previously added */
> > + device_remove_file(dev, &dev_attr_link_10);
> > + device_remove_file(dev, &dev_attr_link_100);
> > + device_remove_file(dev, &dev_attr_link_1000);
> > + device_remove_file(dev, &dev_attr_link_2500);
> > + device_remove_file(dev, &dev_attr_link_5000);
> > + device_remove_file(dev, &dev_attr_link_10000);
>
> Noooooo!
>
> > + ret = __ethtool_get_link_ksettings(trigger_data->net_dev, &cmd);
> > + if (ret)
> > + return ret;
> > +
> > + /* Add only supported entry */
> > + if (test_bit(ETHTOOL_LINK_MODE_10baseT_Half_BIT, cmd.link_modes.supported) ||
> > + test_bit(ETHTOOL_LINK_MODE_10baseT_Full_BIT, cmd.link_modes.supported))
> > + device_create_file(dev, &dev_attr_link_10);
> > +
> > + if (test_bit(ETHTOOL_LINK_MODE_100baseT_Half_BIT, cmd.link_modes.supported) ||
> > + test_bit(ETHTOOL_LINK_MODE_100baseT_Full_BIT, cmd.link_modes.supported))
> > + device_create_file(dev, &dev_attr_link_100);
> > +
> > + if (test_bit(ETHTOOL_LINK_MODE_1000baseT_Half_BIT, cmd.link_modes.supported) ||
> > + test_bit(ETHTOOL_LINK_MODE_1000baseT_Full_BIT, cmd.link_modes.supported))
> > + device_create_file(dev, &dev_attr_link_1000);
> > +
> > + if (test_bit(ETHTOOL_LINK_MODE_2500baseT_Full_BIT, cmd.link_modes.supported))
> > + device_create_file(dev, &dev_attr_link_2500);
> > +
> > + if (test_bit(ETHTOOL_LINK_MODE_5000baseT_Full_BIT, cmd.link_modes.supported))
> > + device_create_file(dev, &dev_attr_link_5000);
> > +
> > + if (test_bit(ETHTOOL_LINK_MODE_10000baseT_Full_BIT, cmd.link_modes.supported))
> > + device_create_file(dev, &dev_attr_link_10000);
> > +
> > + return 0;
> > +}
>
> This should be done via the is_visible sysfs attribute_group method.
>
> Also documentation for the link_* files should be changed so that it
> says that the files are present only if those speeds are available.
>
Just working on doing this change!
Just need to do some test on how is_visible behave with changing the
device_name.
--
Ansuel