Hi,
Werner Almesberger wrote:
Michael Trimarchi wrote:
->ioctl_start ..... by a process, your lock is not acquire
<--- in the middle you acquire the hardware lock and don't made the
hw_unvalaible
---> release the lock at the end
-> call ioctl_wext crash
What I mean is something like this:
ar6000_avail_ev()
{
up_write(&sem);
}
ar6000_unavail_ev()
{
down_write(&sem);
}
ar6000_do_ioctl()
{
...
}
ar6000_ioctl()
{
if (!down_read_trylock(&sem))
return -EIO;
ar6000_do_ioctl();
up_read(&sem);
}
If the unavailability event arrives while we're still in the ioctl,
the down_write just blocks until the ioctl ends and releases the
read lock.
- Werner
Ok, it's the same with one lock. Sorry, for the delay, seems to be a
better solution.
Michael
