Loadable Kernel Modules · The Device Driver Model

1. Demo: cd demo/hello; make; ../../guest/run_guest.sh . then insmod, dmesg, lsmod, rmmod
2. TODO 1 in sysprog_proc_starter.c: module_param(owner, charp, 0444)
3. TODO 2: four seq_printf lines: owner, reads, loaded for, reader
4. TODO 3: proc_create_single in init; NULL → return -ENOMEM
5. TODO 4: remove_proc_entry in exit. Test: insmod owner=…, cat twice, rmmod, cat

the guest, with the finished lab (real)

guest 6.8.0-146-generic: your modules are in /mod:
$ insmod sysprog_proc.ko owner=Mahbub
$ cat /proc/sysprog
owner:      Mahbub
reads:      1
loaded for: 15 jiffies = 15 ms (HZ = 1000)
reader:     cat (pid 136)
$ rmmod sysprog_proc
$ dmesg | tail -2
sysprog: /proc/sysprog created (owner=Mahbub)
sysprog: /proc/sysprog removed after 2 reads

Forget TODO 4 and the first cat after rmmod runs code that is gone: an oops. In the VM that is a lesson; on the PC it would be a crash.

1. cd lab14_2; make; ../guest/run_guest.sh . then insmod sp_dev.ko and sp_drv_starter.ko: no probe (why?)
2. TODO 4 first: the driver's .name must match the device's name
3. TODO 3: probe: devm_kzalloc the state, mutex_init, value = 7, set_drvdata
4. TODO 1, 2: value_show with sysfs_emit; value_store with kstrtoint
5. Test: cat, echo 42, echo abc, unbind, bind, rmmod in both orders

the starter, before TODO 4 (real)

$ insmod sp_dev.ko; insmod sp_drv_starter.ko
$ ls /sys/bus/platform/drivers
ac  acpi-fan  …  syscon
sysprog-TODO  tpm_tis  …
$ ls /sys/devices/platform/sysprog.0
driver_override  power  uevent
modalias  subsystem          # no driver!
# after the lab:
$ cat /sys/devices/platform/sysprog.0/value
7
$ echo 42 > …/value; cat …/value
42

The starter loads but never binds (its driver is called sysprog-TODO). Finding out why, with ls /sys/bus/platform/drivers and dmesg, is part of the lab.