Messages in this thread | | | Date | Tue, 19 Jun 2012 10:14:47 +0100 | From | Alan Cox <> | Subject | Re: [RFC PATCH] pch_uart: Add eg20t_port lock field, avoid recursive spinlocks |
| |
On Mon, 18 Jun 2012 14:41:46 -0700 Darren Hart <dvhart@linux.intel.com> wrote:
> > > On 06/05/2012 04:48 PM, Tomoya MORINAGA wrote: > > On Wed, Jun 6, 2012 at 7:07 AM, Darren Hart <dvhart@linux.intel.com> wrote: > >> Are there still concerns about the additional lock? I'll resend V2 > >> tomorrow with the single whitespace fix if I don't hear anything back today. > > > > I understand your saying. Looks good. > > However, I am not expert of linux-uart core system. > > So, I'd like UART maintainer to give us your opinion. > > Greg, Alan, > > any concerns with the locking approach I've adopted in the patch?
Only the one I noted in my reply the first time around which is that you can't permit tty->low_latency=1 unless your tty receive path is not an IRQ path. From a locking point of view the change makes sense anyway.
Going back over it your console locking also needs care - an oops or printk within the areas the private lock covers will hang the box. That should also probably be a trylock style lock as with the other lock on that path
Alan
| |