how to create kernel thread in 2.6 - ??

[vlyamtse]

i am trying to find some info on how to create/ destroy kernel thread. There's enough examples for it, including samples from "Linux Device Drivers" book. The problem is: all describe 2.4 kernel and api is changed pretty dramatically for 2.6. I've been trying to hack something together, using bits and pieces from different kernel groups

void thread_func( struct work_struct *arg)
{
int i;
wait_queue_head_t wq;
printk("<1>::Inside Thread Function::\n");
init_waitqueue_head(&wq);
for (i=0; i < 100; i++) {
interruptible_sleep_on_timeout(&wq, HZ);
printk("HELLO THREAD ITERATION %d...\n", i);

}
}

When i try to create it in init_module:
...
work_struct ws;
INIT_WORK(&ws, thread_func);
/* and schedule it for execution */
schedule_work(&ws);

The thread is created ( i see "HELLO THREAD" messages printed in dmesg log) but init_module
never return control - the "insmod ./hello.ko" runs forever.
Can someone points me what the problem might be?
Also, is there some tutorials on LKM development for 2.6 kernels?
Thank you,

[coopstah13]

3rd edition of this book covers 2.6 kernel. Fortunately it has been open sourced and is available online in PDF format here.

[vlyamtse]

O'Reilly 3-rd edition is still outdated, For example, the book:

INIT_WORK(struct work_struct *work, void (*function)(void *), void *data);
In reality, current 2.6 kernel defines INIT_WORK with 2 arguments

INIT_WORK(struct work_struct *work, void (*function)(work_struct *));

[coopstah13]

As it says, it is current as of 2.6.10 kernel. Your mileage will vary, the API's are constantly changing, you will have to scour some kernel mailing lists.

[vlyamtse]

I did checking on mailing lists. Found plenty of mails complaining about that change.
Did not found anything explaining how to use it ( INIT_WORK) in new context, specifically,
how to pass the data to thread function expecting work_struct* as argument.
Have i been checking the wrong mailing lists?

[coopstah13]

You could try asking on the mailing list.

[vlyamtse]

Here's another thread issue: looks like the thread "sleep" blocks hardware interrupts. Here's simple thread function that makes 10 iterations with going to sleep for some interval of time. While thread function is running, the system is "loosing" input devices, e.g. keyboard and mouse
become nonfunctional. I wonder what the reason is. The "sleep" task is specified as interruptible, there should be no problems with hardware interrupts.
I also tried wait_event_interruptible_timeout - with the same result.
I am missing something ??


void thread_func(void *ptr)
{
int i;
printk("Hello kthread started\n"winking smiley;
for (i=0; i < 10; i++)
{
printk("kthread sleep iteration %d\n", i);
set_current_state(TASK_INTERRUPTIBLE);
schedule_timeout (4*HZ);
..
}


}
DECLARE_WORK(work, thread_func, (void*) NULL);
static int hello_init(void)
{
printk("Hello, world\n"winking smiley;
schedule_work(&work);
return 0;
}

[sandyrmk]

Hi vlyamtse,

I do face the same problem with INIT_WORK accepting only two arguments and I need to pass data to my thread function. Do you have any idea on how to pass the data to thread function expecting work_struct* as argument.

Thank you in advance,

[oz]

Hello and welcome, sandyrmk!

This thread is two years old so I'm going to lock it down, but please feel free to start a fresh thread of your own if you are having any problems with Linux.

Thank you.