Re: [PATCH v2] mm/page_isolation: fix a deadlock with printk()

[Petr Mladek]

On Wed 2019-10-09 10:46:14, Qian Cai wrote:
> On Wed, 2019-10-09 at 16:24 +0200, Petr Mladek wrote:
> > On Wed 2019-10-09 09:43:13, Qian Cai wrote:
> > > On Wed, 2019-10-09 at 15:27 +0200, Michal Hocko wrote:
> > > > On Wed 09-10-19 09:06:42, Qian Cai wrote:
> > > > [...]
> > > > > https://lore.kernel.org/linux-mm/1570460350.5576.290.camel@xxxxxx/
> > > > > 
> > > > > [  297.425964] -> #1 (&port_lock_key){-.-.}:
> > > > > [  297.425967]        __lock_acquire+0x5b3/0xb40
> > > > > [  297.425967]        lock_acquire+0x126/0x280
> > > > > [  297.425968]        _raw_spin_lock_irqsave+0x3a/0x50
> > > > > [  297.425969]        serial8250_console_write+0x3e4/0x450
> > > > > [  297.425970]        univ8250_console_write+0x4b/0x60
> > > > > [  297.425970]        console_unlock+0x501/0x750
> > > > > [  297.425971]        vprintk_emit+0x10d/0x340
> > > > > [  297.425972]        vprintk_default+0x1f/0x30
> > > > > [  297.425972]        vprintk_func+0x44/0xd4
> > > > > [  297.425973]        printk+0x9f/0xc5
> > > > > [  297.425974]        register_console+0x39c/0x520
> > > > > [  297.425975]        univ8250_console_init+0x23/0x2d
> > > > > [  297.425975]        console_init+0x338/0x4cd
> > > > > [  297.425976]        start_kernel+0x534/0x724
> > > > > [  297.425977]        x86_64_start_reservations+0x24/0x26
> > > > > [  297.425977]        x86_64_start_kernel+0xf4/0xfb
> > > > > [  297.425978]        secondary_startup_64+0xb6/0xc0
> > > > > 
> > > > > where the report again show the early boot call trace for the locking
> > > > > dependency,
> > > > > 
> > > > > console_owner --> port_lock_key
> > > > > 
> > > > > but that dependency clearly not only happen in the early boot.
> > > > 
> > > > Can you provide an example of the runtime dependency without any early
> > > > boot artifacts? Because this discussion really doens't make much sense
> > > > without a clear example of a _real_ lockdep report that is not a false
> > > > possitive. All of them so far have been concluded to be false possitive
> > > > AFAIU.
> > > 
> > > An obvious one is in the above link. Just replace the trace in #1 above with
> > > printk() from anywhere, i.e., just ignore the early boot calls there as they are
> > >  not important.
> > > 
> > > printk()
> > >   console_unlock()
> > >     console_lock_spinning_enable() --> console_owner_lock
> > >   call_console_drivers()
> > >     serial8250_console_write() --> port->lock
> > 
> > Please, find the location where this really happens and then suggests
> > how the real deadlock could get fixed. So far, we have seen only
> > false positives and theoretical scenarios.
> 
> Now the bar is higher again. You are now asking me to actually trigger this
> potential deadlock live. I am probably better off buying some lottery tickets
> then if I could be that lucky.

No, we just do not want to comlicate the code too much just to hide
false positives from lockdep.

I do not ask you to reproduce the deadlock. I ask you to find
a code path where the deadlock might really happen. It seems
that you actually found one in the tty code in the other mail.

Best Regards,
Petr