Why does this device_create() call not create a /dev/ entry? - driver

Question

I'm porting platform driver code to a PCIe variant and I don't understand why I'm not getting a /dev/ entry to show up. The platform driver code that has been modified:
static dev_t first;
static struct class * class;
ATTRIBUTE_GROUPS(my);
static int __init my_pci_init(void)
{
int ret;
/* Create a class entry in sysfs */
if ((class = class_create(THIS_MODULE, "test_driver")) == NULL) {
pr_err("Couldn't create 'struct class' structure.");
ret = -ENODEV;
goto exit;
}
class->dev_groups = my_groups;
/* Create the /dev/ file system entry */
/* return value ignored: there's a 'struct class' to 'struct device' mapping */
if (device_create(class, NULL, first, NULL, KBUILD_MODNAME) == NULL) {
pr_err("Couldn't create entry in '/dev/' file system.");
ret = -ENODEV;
goto exit;
} else {
pr_info("Created a /dev/ entry.");
}
if ((ret = pci_register_driver(&pci_driver)) < 0) {
pr_err("Couldn't register pci driver.");
}
exit:
if (ret < 0) {
my_pci_exit();
pr_err(" ret = %d", ret);
}
return ret;
}
module_init(my_pci_init);
If the module name is 'cz_tdm', I was hoping the above code would create an entry /dev/cz_tdm. At least it did when I was compiling this as a platform driver.
The driver enumerates just fine, an output of lspci shows that the driver was loaded and perusing the sysfs shows that all my attributes in /sys/devices/virtual/... are where I'd expect them to be.
What gives?

Answer 1

Whoops.
Because it's not supposed too. An overzealous deletion of code ripped out this necessary element:
/* Add the char device to the system. */
cdev_init(&cdev, &fops);
if ((ret = cdev_add(&cdev, first, DEV_MINOR_NUMBER_COUNT)) < 0) {
pr_err("Couldn't add device to system: %d", ret);
ret = -ENODEV;
goto exit;
}