Re: [PATCH 2.6.21-rc1] ibmebus: Support dynamic addition and removal of adapters

[Sylvain Munaut <tnt@246tNt.com>]

Hoang-Nam Nguyen wrote:
> Additionally, the uevent interface is now implemented in the driver.
>   
Mmmh, I posted a patch that added a common uevent interface for all
of_device based
bus. And that kinda clash with this one.

I think it's a much cleaner approach to make it as common as possible.
But as
"ibm,loc_code" is local to this bus it's not handled by my code. Is it
necessary
to match the correct driver ?

By the way you generate your uevent string is wrong imho:
 - it doesn't account for the fact that there can be multiple strings in
"compatible".
 - it can be a lot nicer if you use the helper add_uevent_var to fill
the event.
 - you generate a string "ibmebus:T%s:S%s:L%s" but did you just invent that
string ? I see no support for the "ibmebus" marker in mod_devicetable.h, nor
in file2alias.c nor any where else for that matter.


    Sylvain


>  
> +static int ibmebus_uevent(struct device *dev, char **envp, int num_envp,
> +			char *buffer, int buffer_size)
> +{
> +	const struct ibmebus_dev *ebus_dev = to_ibmebus_dev(dev);
> +	char *name, *cp, *loc_code;
> +	int length;
> +
> +	if (!num_envp)
> +		return -ENOMEM;
> +
> +	if (!ebus_dev->ofdev.node)
> +		return -ENODEV;
> +
> +	name = (char *)get_property(ebus_dev->ofdev.node, "name", NULL);
> +	cp = (char *)get_property(ebus_dev->ofdev.node, "compatible", NULL);
> +	loc_code = (char *)get_property(ebus_dev->ofdev.node,
> +					"ibm,loc-code", NULL);
> +	if (!(name && cp && loc_code))
> +		return -ENODEV;
> +
> +	envp[0] = buffer;
> +	length = scnprintf(buffer, buffer_size,
> +			   "MODALIAS=ibmebus:T%s:S%s:L%s",
> +			   name, cp, loc_code);
> +	if (buffer_size - length <= 0)
> +		return -ENOMEM;
> +	envp[1] = NULL;
> +
> +	return 0;
> +}
> +
>