Why dev.bus is NULL in my device?

i'm trying to understand how linux device/driver model works and to do this i've written a little module. This module is simple, retrieves a pointer to a struct net_device (let's call it netdev) by the function dev_get_by_name(&init_net, "eth0"). Why the value of netdev->dev.bus is NULL? Should that pointer represent the bus_type structure on which my device is attached? The field netdev->parent->bus is however not NULL but it should represent the bus for eth开发者_JAVA百科 controller...any explanation?

---

This is because your eth device, or better said its device "object" in the kernel, is not a bus and thus its bus value is left unitialized. But its parent device usually is on a bus and it is sufficient that the parent device knows the bus it is on, since both device eventually are linked during the driver initialization.

Let's have a look at an example: here is what I have in sysfs for my eth0 device (notice the device field):

$ ll /sys/class/net/eth0/
total 0
-r--r--r-- 1 root root 4096 May 20 11:10 address
-r--r--r-- 1 root root 4096 May 20 11:10 addr_len
-r--r--r-- 1 root root 4096 May 20 11:10 broadcast
-r--r--r-- 1 root root 4096 May 20 11:10 carrier
lrwxrwxrwx 1 root root    0 May 20 11:10 device -> ../../../devices/pci0000:00/0000:00:19.0
-r--r--r-- 1 root root 4096 May 20 11:10 dev_id
-r--r--r-- 1 root root 4096 May 20 11:10 dormant
-r--r--r-- 1 root root 4096 May 20 11:10 features
-rw-r--r-- 1 root root 4096 May 20 11:10 flags
-rw-r--r-- 1 root root 4096 May 20 11:10 ifalias
-r--r--r-- 1 root root 4096 May 20 11:10 ifindex
-r--r--r-- 1 root root 4096 May 20 11:10 iflink
-r--r--r-- 1 root root 4096 May 20 11:10 link_mode
-rw-r--r-- 1 root root 4096 May 20 11:10 mtu
-r--r--r-- 1 root root 4096 May 20 11:10 operstate
drwxr-xr-x 2 root root    0 May 20 11:10 power
drwxr-xr-x 2 root root    0 May 20 11:10 statistics
lrwxrwxrwx 1 root root    0 May 20 11:10 subsystem -> ../../net
-rw-r--r-- 1 root root 4096 May 20 11:10 tx_queue_len
-r--r--r-- 1 root root 4096 May 20 11:10 type
-rw-r--r-- 1 root root 4096 May 20 11:10 uevent

The link for the device is created from this code from the driver probe function, where netdev is the network device, and pdev the associated PCI device:

SET_NETDEV_DEV(netdev, &pdev->dev);

Which according to the documentation is:

/* Set the sysfs physical device reference for the network logical device
 * if set prior to registration will cause a symlink during initialization.
 */
#define SET_NETDEV_DEV(net, pdev)   ((net)->dev.parent = (pdev))

And here is what I have in the corresponding PCI device, that was set by SET_NETDEV_DEV (where you can notice the bus field):

$ ll /sys/devices/pci0000\:00/0000\:00\:19.0/
total 0
-rw-r--r-- 1 root root   4096 May 20 11:54 broken_parity_status
lrwxrwxrwx 1 root root      0 May 20 11:22 bus -> ../../../bus/pci
-r--r--r-- 1 root root   4096 May 20 11:07 class
-rw-r--r-- 1 root root    256 May 20 11:22 config
-r--r--r-- 1 root root   4096 May 20 11:54 device
lrwxrwxrwx 1 root root      0 May 20 11:22 driver -> ../../../bus/pci/drivers/e1000e
-rw------- 1 root root   4096 May 20 11:22 enable
-r--r--r-- 1 root root   4096 May 20 11:07 irq
-r--r--r-- 1 root root   4096 May 20 11:54 local_cpulist
-r--r--r-- 1 root root   4096 May 20 11:07 local_cpus
-r--r--r-- 1 root root   4096 May 20 11:22 modalias
-rw-r--r-- 1 root root   4096 May 20 11:22 msi_bus
lrwxrwxrwx 1 root root      0 May 20 11:22 net:eth0 -> ../../../class/net/eth0
drwxr-xr-x 2 root root      0 May 20 11:11 power
-r--r--r-- 1 root root   4096 May 20 11:22 resource
-rw------- 1 root root 131072 May 20 11:22 resource0
-rw------- 1 root root   4096 May 20 11:22 resource1
-rw------- 1 root root     32 May 20 11:22 resource2
lrwxrwxrwx 1 root root      0 May 20 11:22 subsystem -> ../../../bus/pci
-r--r--r-- 1 root root   4096 May 20 11:22 subsystem_device
-r--r--r-- 1 root root   4096 May 20 11:22 subsystem_vendor
-rw-r--r-- 1 root root   4096 May 20 11:22 uevent
-r--r--r-- 1 root root   4096 May 20 11:22 vendor

I hope this clarifies the situtation.