Henrique Montenegro [off-list ref] writes:

Hello list,

I am reading through the e1000 driver and trying to figure out how the
probe function on it gets called.

The driver initialization function calls pci_register_driver:

----------------------------------------------------------------------------->8
static struct pci_driver e1000_driver = {
.name     = e1000_driver_name,
.id_table = e1000_pci_tbl,
.probe    = e1000_probe,
.remove   = e1000_remove,
    // ...
};

static int __init e1000_init_module(void)
{
    // ...
ret = pci_register_driver(&e1000_driver);
    // ...
}
----------------------------------------------------------------------------->8

And pci_register_driver is defined as (on linux/pci.h):

----------------------------------------------------------------------------->8
#define pci_register_driver(driver)             \
         __pci_register_driver(driver, THIS_MODULE, KBUILD_MODNAME)
----------------------------------------------------------------------------->8

Function __pci_register_driver is defined as (drivers/pci/pci-driver.c):

----------------------------------------------------------------------------->8
int __pci_register_driver(struct pci_driver *drv, struct module *owner,
                          const char *mod_name)
{
        /* initialize common driver fields */
        drv->driver.name = drv->name;
        drv->driver.bus = &pci_bus_type;
        drv->driver.owner = owner;
        drv->driver.mod_name = mod_name;

        spin_lock_init(&drv->dynids.lock);
        INIT_LIST_HEAD(&drv->dynids.list);

        /* register with core */
        return driver_register(&drv->driver);
}
EXPORT_SYMBOL(__pci_register_driver);
----------------------------------------------------------------------------->8

This is the point where I am getting lost. I can't figure out how the Kernel
will know about the functions defined in the e1000_driver struct mentioned
before, since it does not seem to pass a reference to it anywhere.

How does the kernel know where the probe function for this module is in this
case? To be honest, for any driver that calls pci_register_driver, how will
the
kernel know where the probe function is since it does not seem like it is
being passed to driver_register?

The magic is in the 'drv->driver.bus = &pci_bus_type;' assigment.  This
is where the driver core will look for functions knowing how to handle
this specific driver.  See Documentation/driver-model/bus.txt etc

Look at the defintion of pci_bus_type in  drivers/pci/pci-driver.c :

struct bus_type pci_bus_type = {
        .name           = "pci",
        .match          = pci_bus_match,
        .uevent         = pci_uevent,
        .probe          = pci_device_probe,
        .remove         = pci_device_remove,
        .shutdown       = pci_device_shutdown,
        .dev_groups     = pci_dev_groups,
        .bus_groups     = pci_bus_groups,
        .drv_groups     = pci_drv_groups,
        .pm             = PCI_PM_OPS_PTR,
};
EXPORT_SYMBOL(pci_bus_type);

And then look at the different callbacks.  These explain how the generic
&drv->driver above is turned back into a pci_driver on probing:

static int pci_device_probe(struct device *dev)
{
        int error;
        struct pci_dev *pci_dev = to_pci_dev(dev);
        struct pci_driver *drv = to_pci_driver(dev->driver);


to_pci_dev() and to_pci_driver() are just macros simplifying the usual
container_of trick.  From include/linux/pci.h :

#define to_pci_dev(n) container_of(n, struct pci_dev, dev)
..
#define to_pci_driver(drv) container_of(drv, struct pci_driver, driver)



Hope this helps.


Bj?rn