On 02/01/2012 12:28 PM, Dmitry Antipov wrote:
I'm writing a kernel module which creates a substantial amount of kernel threads. After dropping the real stuff, the module skeleton is:
#include <linux/kernel.h> #include <linux/sched.h> #include <linux/kthread.h> #include <linux/module.h> #include <linux/slab.h> #include <linux/ktime.h>
MODULE_LICENSE("GPL");
static int nrthreads = 128; module_param(nrthreads, int, 0644);
static int loopcount = 1024; module_param(loopcount, int, 0644);
static struct task_struct **threads; static struct completion done; static atomic_t nrunning;
static int test(void *unused) { int i; ktime_t expires = ktime_set(0, NSEC_PER_MSEC);
for (i = 0; !kthread_should_stop() && i < loopcount; i++) schedule_hrtimeout_range(&expires, 50000, HRTIMER_MODE_REL);
if (atomic_dec_and_test(&nrunning)) complete(&done); return 0; }
static int __init testmod_init(void) { int i, j, err = 0;
atomic_set(&nrunning, 0); init_completion(&done);
threads = kmalloc(nrthreads * sizeof(struct task_struct *), GFP_KERNEL); if (!threads) return -ENOMEM;
for (i = 0; i < nrthreads; i++) { threads[i] = kthread_run(test, NULL, "test/%d", i); if (IS_ERR(threads[i])) { err = PTR_ERR(threads[i]); for (j = 0; j < i; j++) kthread_stop(threads[j]); kfree(threads); return err; } atomic_inc(&nrunning); } return 0; }
static void __exit testmod_exit(void) { wait_for_completion(&done); kfree(threads); }
module_init(testmod_init); module_exit(testmod_exit);
For the most of the cases, it works as expected, at least from 8 to 128 threads. But if I try 'insmod testmod.ko && rmmod testmod', it's possible to catch a very rare crash:
IMO, you have a race condition with nrunning. What guarantee do you have atomic_dec_and_test is called after atomic_inc ?
Maybe you can try by removing atomic_inc and do atomic_set(&nrunning, nrthreads) ?