but, does the lock_bh, and lock_irqsave's share the counter ?
(i.e. both use local_bh_disable/enable)
I'm not too familiar with the details, but from what I gathered, that's what would happen.
but since I'm guessing, it probably is incorrect.
On 4/27/06, Daniel Wu <firstname.lastname@example.org> wrote:
> > On a possibly related note: ieee80211_new_state is sometimes called
> > from softirq context. It uses spin_lock instead of spin_lock_bh. Is
> > this a potential problem?
> That may be true, but as long as the ieee80211_new_state is called inbetween
> an irq_save block, calling the spin_lock_bh/unlock_bh is bad since unlock_bh
> will cause enable_bh to be called which it should be in irq_save block.
Are you sure? spin_lock_bh seems to use a counter rather than an on/off flag.