Re: [RFC][PATCH v5 1/2] printk: Make printk() completely async
From: Jan Kara
Date: Mon Mar 21 2016 - 11:32:45 EST
On Mon 21-03-16 23:58:32, Sergey Senozhatsky wrote:
> Hello Jan,
> On (03/21/16 15:32), Jan Kara wrote:
> > > we have 2 spin locks in vprintk_emit() -- logbuf_lock and sem->lock. and N
> > > CPUs can concurrently lockup on those two locks, which already makes a
> > > single static pointer in spiun_dump() questionable.
> > >
> > > logbug_lock *theoretically* can detect and handle recursive printk()s,
> > > there is no way to catch sem->lock spin_dump() at the moment (but that's
> > > not the point).
> > >
> > > there are 2 new spin locks in vprintk_emit() -- p->pi_lock and rq->lock.
> > Actually, this is not true. These locks are already in vprintk_emit() via
> > the up(&console_sem) call from console_unlock() since up() can call
> > wake_up() which needs the same locks as wake_up_process().
> true. I meant new locks (which come with printk_kthread). the already
> existing locks and problems were not addressed.
> > And by calling wake_up_process() under logbuf_lock, you actually introduce
> > recursion issues for printk_deferred() messages which are supposed to be
> > working from under rq->lock and similar. So I think you have to keep this
> > section outside of logbuf_lock.
> hm, in_sched (printk_deferred()) messages are printed by
> irq work->wake_up_klogd_work_func(), not by wake_up_process()
> from vprintk_emit(). or am I missing something?
Think of following:
vprintk_emit() -> recursion on logbuf_lock
Previously scheduler code was allowed to call printk_deferred() wherever it
So we are not supposed to call into the scheduler from under logbuf_lock...
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR