Re: [PATCH v6 3/5] Implement modalias sysfs attribute for modules

[Greg Kroah-Hartman]

On Fri, Dec 02, 2022 at 04:47:42PM -0600, Allen Webb wrote:
> When the modalias attribute is read, invoke a subsystem-specific
> callback for each driver registered by the specific module.
> 
> The intent of the new modalias attribute is to expose the
> match-id-based modaliases to userspace for builtin and loaded kernel
> modules.
> 
> Signed-off-by: Allen Webb <allenwebb@xxxxxxxxxx>
> ---
>  include/linux/device/bus.h |  7 +++++
>  kernel/module/sysfs.c      | 57 +++++++++++++++++++++++++++++++++++++-
>  2 files changed, 63 insertions(+), 1 deletion(-)
> 
> diff --git a/include/linux/device/bus.h b/include/linux/device/bus.h
> index 82a5583437099..cce0bedec63d9 100644
> --- a/include/linux/device/bus.h
> +++ b/include/linux/device/bus.h
> @@ -61,6 +61,10 @@ struct fwnode_handle;
>   *			this bus.
>   * @dma_cleanup:	Called to cleanup DMA configuration on a device on
>   *			this bus.
> + * @drv_to_modalias:    Called to convert the matching IDs in a
> + *                      struct device_driver to their corresponding modaliases.
> + *                      Note that the struct device_driver is expected to belong
> + *                      to this bus.

If the driver was not part of this bus, that just wouldn't work at all
so I don't think you need to say this.

And what is the format here?  New lines?  structures?


>   * @pm:		Power management operations of this bus, callback the specific
>   *		device driver's pm-ops.
>   * @iommu_ops:  IOMMU specific operations for this bus, used to attach IOMMU
> @@ -107,6 +111,9 @@ struct bus_type {
>  	int (*dma_configure)(struct device *dev);
>  	void (*dma_cleanup)(struct device *dev);
>  
> +	ssize_t (*drv_to_modalias)(struct device_driver *drv, char *buf,
> +				   size_t count);
> +
>  	const struct dev_pm_ops *pm;
>  
>  	const struct iommu_ops *iommu_ops;
> diff --git a/kernel/module/sysfs.c b/kernel/module/sysfs.c
> index 8dafec7455fbe..651c677c4ab96 100644
> --- a/kernel/module/sysfs.c
> +++ b/kernel/module/sysfs.c
> @@ -5,6 +5,8 @@
>   * Copyright (C) 2008 Rusty Russell
>   */
>  
> +#include <linux/device/bus.h>
> +#include <linux/device/driver.h>

That feels wrong, modules shouldn't care about busses or drivers.

Why can't this all be in the driver core instead?

>  #include <linux/module.h>
>  #include <linux/kernel.h>
>  #include <linux/fs.h>
> @@ -240,11 +242,64 @@ static inline void add_notes_attrs(struct module *mod, const struct load_info *i
>  static inline void remove_notes_attrs(struct module *mod) { }
>  #endif /* CONFIG_KALLSYMS */
>  
> +/* Track of the buffer and module identity in callbacks when walking the list of
> + * drivers for each bus.
> + */
> +struct modalias_bus_print_state {
> +	struct module_kobject *mk;
> +	char *buf;
> +	size_t count;
> +	ssize_t len;
> +};
> +
> +static int print_modalias_for_drv(struct device_driver *drv, void *p)
> +{
> +	struct modalias_bus_print_state *s = p;
> +	struct module_kobject *mk = s->mk;
> +	ssize_t len;
> +	/* Skip drivers that do not match this module. */

Always use checkpatch.pl on your changes before sening them out :(

> +	if (mk->mod) {
> +		if (mk->mod != drv->owner)
> +			return 0;
> +	} else if (!mk->kobj.name || !drv->mod_name ||
> +		   strcmp(mk->kobj.name, drv->mod_name))
> +		return 0;
> +
> +	if (drv->bus && drv->bus->drv_to_modalias) {
> +		len = drv->bus->drv_to_modalias(drv, s->buf + s->len,
> +						s->count - s->len);
> +		if (len < 0)
> +			return len;
> +		s->len += len;
> +	}
> +	return 0;
> +}
> +
> +static int print_modalias_for_bus(struct bus_type *type, void *p)
> +{
> +	return bus_for_each_drv(type, NULL, p, print_modalias_for_drv);
> +}
> +
>  static ssize_t module_modalias_read(struct file *filp, struct kobject *kobj,
>  				    struct bin_attribute *bin_attr,
>  				    char *buf, loff_t pos, size_t count)
>  {
> -	return 0;
> +	struct module_kobject *mk = container_of(kobj, struct module_kobject,
> +						 kobj);
> +	struct modalias_bus_print_state state = {mk, buf, count, 0};

You allocate this on the stack?

> +	int error = 0;

No need to initialize this.

> +
> +	if (pos != 0)
> +		return -EINVAL;

Why?

> +
> +	error = bus_for_each(&state, print_modalias_for_bus);

So for every module attribute, we walk all busses in the system?  Why
not the bus that this module has a driver for instead?

thanks,

greg k-h